© 1996 by Chris Woodhouse, all rights reserved

Image Capture

 

Imaging Paths

Different ways from image capture to final print

There are numerous photographic methods to get from an appealing subject to a fascinating image. To the observer, none of them are of any consequence, because to him or her, the final image is the only reference. If the image is poor to begin with, there is no need to explore it any further. If it is a striking image, however, who cares how it was made? Photographers, on the other hand, do care, because technical expertise and craftsmanship are part of their creative process, and they are always interested in opportunities to explore new techniques and improve their skill.

This edition of the book is exclusively concerned with monochrome imaging. The previous edition was further restricted to dominantly cover analog monochrome photography. This concentrated our efforts on an imaging path involving analog, film-based cameras for image capture, and traditional darkroom work and silver-gelatin prints for image output (fig.1). To us, this is a logical preference, because we trust an analog imaging path to satisfy our high standards of fine-art printing. Even so, there are other methods to create eye-catching images, and this edition of the book gives us the opportunity to explore them in addition to traditional methods.

Fig.1 is not a complete list of all imaging possibilities, but it illustrates that many other alternatives exist. For example, digital cameras and scanners are the direct and indirect gateways to the fascinating world of digital imaging. They either replace the analog film as an image-capturing medium altogether, or they complement the analog input by scanning the film emulsion pixel-by-pixel. After scanning, the analog image information is converted to digital data, which is then available for a more flexible computer-based image manipulation. Fundamentally, there is little difference between image manipulation using computer software and tonal corrections in the darkroom. Both are effective tools to optimize the image, highlight what is important and suppress or eliminate what is not. Nevertheless, in some cases, digital image manipulation offers more flexibility and additional opportunities for creative expression. For these cases, it makes sense to temporarily switch from analog to digital in order to gain an additional set of tools for image improvement and optimization.

fig.1       There are many possible imaging paths to get from the subject to the print, and combining analog and digital elements can help to optimize image quality.

Unfortunately, many photographers treat this switch like a one-way ticket. Once in the digital domain, they remain there without realizing that an imaging path using analog and digital elements can be more beneficial than a pure analog or digital imaging path alone. They are missing the opportunity to cherry-pick their way along the imaging chain to create the best image possible.

To us, the beauty of a silver-gelatin print is something very special and second to none. Therefore, we continue to concentrate heavily on traditional photographic techniques in order to create an analog print. However, starting with this edition, we use and explain analog and digital capture mechanisms, within their limitations, and integrate darkroom and computer-based manipulations to get the best-looking print possible. In other words, as long as the final output of our creative efforts is a traditional silver-gelatin print, we consider any reasonable deviation from the traditional imaging path an alternative.

For example, modern digital cameras capture just enough image information to satisfy the requirements of standard print observation. However, these images can be manipulated and optimized beyond the possibilities of the standard darkroom, using a computer and imaging software. Afterwards, they can be brought back to the analog domain by creating a digital negative with the help of an imagesetter, film writer or digital printer. Back in the darkroom, the digital negative is contact printed onto traditional photographic paper, and a digital image is successfully converted into a quality silver-gelatin print. To capture even more image data and satisfy the requirements of critical print observation, start with a traditional B&W film or print, scan it at high resolution, manipulate the image data on the computer, and bring it back to the darkroom via a digital negative as explained above. Film remains a viable option until further improvements in digital-camera technology are made. Nevertheless, it will always be an option for people who prefer the security of an additional archive and ever-compatible media.

As you can see, this edition of the book expands the imaging path to include the exciting opportunities and the flexibility of digital input and manipulation, conditional on meeting our quality standards. However, the book also remains dedicated to monochrome B&W photography and still considers silver-gelatin paper to be the best output medium for high-quality prints. Digital negatives open the world of digital imaging to fine-art printers without forcing us into compromised image-output alternatives.

 

Sharpness and Depth of Field

About the limits of human vision and image clarity

If your camera is precisely aligned and the lens is focused at a specific subject distance, then all objects at precisely this distance are in focus, and strictly speaking, everything else is out of focus. In reality, however, our eyes have a limited optical resolution, and therefore, objects reasonably close to the focus plane also appear perfectly focused in the final print. This creates a zone of still acceptable focus surrounding the focus plane, and objects within this zone are considered to be in focus, while those outside are out of focus. This zone is called the depth of field.

Limits of Human Vision and Normal Viewing Distance

The limits of human vision differ substantially with the shape and pattern of the object being observed. The eye’s capability to recognize a single line is astonishing. A dark human hair is easily distinguished against a well-illuminated white background at a distance of 10 m (30 feet) or more. This calculates to a visual angle of about 1 arc second (0°00’01"). The eye’s capability to recognize a single point is less impressive. The size of the smallest object, clearly and consistently visible to most people, calculates to a visual angle of about 1 minute of arc (0°01’00").

Neither of these two tests realistically represents what happens during the observation of a photograph, where visual quality is not challenged by the ability of the eye to detect individual image elements but to resolve fine detail. Several line patterns are used in ergonomic studies to support an objective measure for the resolving power of the human eye. With these patterns, resolving power is measured as the capacity of the eye to discriminate closely spaced lines as separate and distinct line images (see fig.1a).

© 2001 by Lynn Radeka, all rights reserved

A commonly agreed result of these studies is that the minimum visual angle at which a line is perceived within a pattern of three bars, separated by spaces of equal width, is about 1 minute of arc (0°01’00"). Beyond these studies, empirical tests have shown that common detail and distinct texture are still visible down to about 20 seconds of arc (0°00’20"), a value that must be considered for critical observation. Finally, other factors, such as image contrast and ambient illumination, significantly influence the minimum visual angle. Use fig.1a-c to find your personal limits, but for the rest of this book, the minimum visual angle of the middle-aged human eye is assumed to be between 20 seconds and 1 minute of arc, which is the range from critical to standard observation, respectively.

fig.1a    The USAF/1951 test pattern is divided into groups with six elements each.

fig.1b    visual angle in arc minutes (at 1 m distance)

fig.1c    test pattern resolution in lp/mm

fig.1a-c You can use the USAF/1951 test pattern to check your personal limits of vision. If applicable, conduct the tests using your prescribed corrective glasses.

1.) Place fig.1a in a well-lit area, and evaluate the test pattern from a fixed distance of 1 m (40 inches). Find the group and element where you can still make out a line pattern, and fig.1b will reveal your minimum visual angle in arc minutes.

2.) Place fig.1a in a well-lit area, and evaluate the test pattern from a distance of approximately 250 mm (10 inches). Find the group and element where you can still make out a line pattern, and fig.1c will reveal your near-vision resolving power in line pairs per millimeter (lp/mm).

You can also use the USAF/1951 test pattern to evaluate the performance of your photographic lenses.

3.) Mount camera and lens onto a tripod, and use a fine-grain film to take a photograph of fig.1a from a distance equal to a known multiple of the focal length (25-100x). Consider the use of a cable release and a flash to reduce camera-shake as much as possible. Inspect the negative with a loupe and find the group and element where you can still make out a line pattern. Identify the accompanying resolution of the test pattern in fig.1c, and multiply that value by the focal-length multiplier to find the actual lens resolution in line pairs per millimeter (lp/mm).

Of course, the minimum visual angle itself does not tell much about the best optical resolution of photographic detail. We must also be aware of the minimum viewing distance to the photograph. Physiological limitations place comfortable near distance vision at about 250 mm (10 inches), and a most critical viewer may be as close as his or her eyes will focus, investigating all areas of the photograph. Aside from photographic competition judges, this is probably the exception, but at this distance, standard human vision resolves 7 lp/mm (line pairs per millimeter), and critical observation senses detail all the way down to 20 lp/mm, which is still well within the resolution limit of photographic paper (60 lp/mm).

In order to keep depth-of-field scales independent of print size, lens and camera manufacturers make the reasonable assumption that for uncropped prints of 8x10 inches or larger, the normal viewing distance is approximately equal to the print diagonal. For an 8x10-print and the standard minimum visual angle of 1 minute of arc, this calculates to a minimum viewing distance of 325 mm and a resolving power of 10 lines/mm or 5 lp/mm. In other words, at this distance and under normal viewing conditions, the human eye cannot separate print detail smaller than 0.2 mm.

To make an 8x10-inch print from the entire 35mm negative requires about an 8.5-times enlargement. Therefore, negative detail is 8.5 times smaller than its respective print detail, and consequently, the maximum print detail of 0.2 mm has an equivalent maximum negative detail of 0.022 mm (see fig.2). Any 35mm-negative detail smaller than 0.022 mm cannot be resolved during print observation and, consequently, does not have to be in focus on the negative to appear resolved in the print.

Although we have used the 8x10-print as an example, our assumption of a fixed relationship between viewing distance and print size is appropriate for all print sizes. Any change to the negative magnification is mathematically compensated for by a change in viewing distance. This conveniently keeps the size of the minimum negative detail, for all full-negative enlargements of a given negative format, consistent.

fig.2       If a print is observed under normal viewing conditions, human vision can detect individual image elements as small as perceived within the minimal visual angle. However, in order to resolve print detail, twice that angle is needed. Making a print from a negative typically requires a certain magnification. The acceptable circle of confusion is smaller than its respective print detail by this factor of magnification. Any negative detail smaller than the acceptable circle of confusion cannot be resolved during the above print observation and, therefore, does not have to be in focus on the negative to appear clear in the print.

fig.3       A point light source is projected by the lens as a cone of light that converges towards the plane of perfect focus at the film plane, where it forms a tiny point. If focused slightly in front of, or behind, the light, it will change to a small blurry circle. As long as the blurry circle is smaller than the minimum negative detail, it will appear as a point when enlarged for printing. The blurry circle is the ‘circle of confusion’.

(illustration based on an original by John B. Williams)

Circle of Confusion

Imagine the following experiment. In a darkened room, point your camera and normal lens towards the lit bulb of a miniature flashlight placed as far away as possible. The pinpoint light is extremely small and has practically no height or width. If you focus the lens on that light, it forms a tiny point on the view screen. However, if you focus slightly in front of, or behind, the light, it will change to a small blurry circle (fig.3). As long as that blurry circle is smaller than the minimum negative detail, it will look like a point when enlarged for printing. The blurry circle is the ‘circle of confusion’.

fig.4       The acceptable circle of confusion for standard and critical viewing conditions depends on the negative format and the optical resolution limits of the eye.

Except for the purpose of close inspection, we assume that the minimum distance from which a print is viewed is about equal to the print diagonal. This assumption allows us to work with one fixed-size circle of confusion per negative format, because print size and viewing distance grow proportionally. Small negative formats require more negative magnification to produce the same size print than larger formats. Consequently, small negative formats need smaller negative detail and smaller circles of confusion than larger formats. If we assume that the entire negative is printed to produce the print, fig.4 gives standard and critical dimensions for the acceptable circle of confusion, and the minimum negative resolutions required to achieve them, for several negative formats.

Depth of Field

The flashlight experiment clearly shows that there is a zone of still acceptable focus surrounding the focus plane, and its size depends on several variables in addition to the circle of confusion. The depth of field increases with subject distance and decreases with focal length. As a result, the longer the focal length or the closer the subject, the shallower the depth of field. In macro photography, the depth of field is often reduced to just a few millimeters. Short focal-length lenses provide more depth of field than long focal-length lenses from the same viewpoint, even when the negative is printed with a higher magnification to render the same scale print.

fig.5a-b  A smaller lens opening permits only the light that is close to the center of the optical axis to reach the film. As a result, the image is dimmer, but the depth of field is increased.

The last significant variable for the depth of field is the lens aperture. Fig.5a-b show how the circle of confusion makes depth of field possible and how the zone increases as the aperture is reduced. In fig.5a, a large aperture limits the depth of field to a relatively small zone. The image circle of a far point is larger than the circle of confusion, and therefore, the point is out of focus. In fig.5b, a smaller lens opening permits only the light that is close to the center of the optical axis to reach the film. As a result, the image is dimmer, but the depth of field is increased.

Closing the aperture by a few stops makes for a significant increase in depth of field. Eventually, the lens aperture is small enough for the depth of field to reach infinity. The tiny aperture of a pinhole camera, often smaller than f/256, produces images approaching infinite depth of field from front to rear.

Quality small- and medium-format lenses have engraved depth-of-field scales as a practical aid for optimal depth-of-field placement or convenient zone focusing. In my experience, many of these scales use a rather optimistic circle of confusion, which makes for an only mediocre depth of field. If you have more stringent requirements, stop the lens one or two stops further down than what the scale suggests, or calculate a personalized depth-of-field table. The equations to calculate the depth of field (dF), and its front (df) and rear (dr) limits, are:

where ‘u’ is the focusing distance, ‘f’ is the focal length, ‘c’ is the circle of confusion, and ‘N’ is the lens aperture in f/stops.

In case the subject magnification is already known or calculated, the equation to determine the depth of field (dF) simplifies to:

where ‘c’ is the circle of confusion, ‘N’ is the lens aperture in f/stops, and ‘m’ is the subject magnification. This equation is adequately accurate for subject magnifications larger than 0.1, which means it can be used for close-up but not landscape photography.

With the help of a spreadsheet and the equations provided here, customized tables for many formats and lenses can be prepared and then kept in the camera bag for future assignments. When performing the computations, be sure to keep units consistent and not to mix imperial and metric units.

Hyperfocal Distance

Maximum depth of field is obtained in any situation through use of the hyperfocal distance. The hyperfocal distance is defined as the minimum focus distance at which the rear depth-of-field limit is equal to infinity. This has the following consequences: If a lens is focused at the hyperfocal distance, the depth of field starts at half the hyperfocal distance and ends at infinity. But, if the lens is focused at infinity, the depth of field extends only from the hyperfocal distance to infinity (fig.6). The hyperfocal distance (dH) is accurately given by:

or simplified, but adequately accurate given by:

where ‘f’ is the focal length, ‘c’ is the circle of confusion, and ‘N’ is the lens aperture in f/stops.

One noteworthy advantage of using the hyperfocal distance is that, once known, the formulae to calculate the front (df) and rear (dr) depth-of-field limits are much simplified to:

fig.6       If a lens is focused at the hyperfocal distance, the depth of field starts at half the hyperfocal distance and ends at infinity. But, if the lens is focused at infinity, the depth of field extends only from the hyperfocal distance to infinity.

where ‘dH’ is the hyperfocal distance and ‘u’ is the focusing distance. These simplified formulae lack the accuracy of the equations on the previous page, but they can be used without hesitation for focus distances greater than 10 times the focal length.

Depth of Focus

As seen in fig.5, similar to the zone of reasonable focus surrounding the focal plane, known as the depth of field, there is an equivalent zone of reasonable focus surrounding the film plane, called the depth of focus (dF’).

As the film image is a scaled version of the subject in front of the camera, the depth of focus is a scaled version of the depth of field (fig.7). The front (df’) and rear (dr’) limits of the depth of focus can be calculated from the front and rear depth-of-field limits by:

fig.7       This illustration demonstrates the relationship between depth of field and depth of focus. Depth of focus increases with the circle of confusion and magnification. It decreases with increasing lens aperture and is at its minimum when the lens is focused at infinity.

(based on an original by Harold M. Merklinger)

where ‘f’ is the focal length, and ‘df’ and ‘dr’ are the front and rear depth-of-field limits around the focal plane.

Depth of focus increases with the circle of confusion and subject magnification. It decreases as the lens aperture increases, and it is at its minimum when the lens is focused at infinity. Alternatively, the total depth of focus (dF’) is, therefore, also given by:

where ‘c’ is the circle of confusion, ‘N’ is the lens aperture in f/stops, and ‘m’ is the subject magnification, but the formula simplifies to:

if the lens is focused at or near infinity, at which point the magnification (m) is insignificantly small and approaching zero.

View camera lenses do not usually feature distance or depth-of-field markings. At first thought, this makes reaching the required depth of field through f/stop estimates impossible, or at least difficult and cumbersome. Nevertheless, since the depth of focus is directly related to the depth of field, this relationship can be used as a reliable alternative when operating a view camera at or near infinity focus.

fig.8a    The depth-of-focus scale and the gauges shown here are based on the standard circle of confusion for several view-camera formats and can be used with any focal length. Make a copy of each for your personal use.

fig.8b    Mount the depth-of-focus scale to the camera, mark the near and far focus positions of the focusing standard on the scale, and use the appropriate gauge to translate the distance between them into the required aperture. Then, move the focusing standard to the optimum focusing position, which is midway between the markings for near and far focus. This way, depth of field will be achieved between the near and far focal planes.

Fig.8a shows a depth-of-focus scale and gauges for several view-camera formats; fig.8b shows one set in operation. Mount the scale to the monorail or the camera bed of your view camera. Focus the camera on the most distant point for which resolution of detail is required and mark the position of the focusing standard to the scale. Then, focus on the nearest point for which resolution of detail is required, mark its position and measure the distance. Use the appropriate depth-of-focus gauge to translate this distance into the minimum aperture necessary, and slide the focusing standard to the optimum focusing position, located midway between the markings for near and far focus. Each gauge is dedicated to a specific film format but can be used with any focal length, because all gauges are designed for near-infinity focus conditions.

Diffraction or Limits of Resolution

In practice, lens resolution is limited by two factors, aberrations and diffraction. The resolution limit, due to optical aberrations, depends on lens design and construction, and aberrations are reduced as the lens is stopped down and the aperture gets smaller. Modern lens designs have minimized aberrations, and between f/8 and f/11, lens resolutions of 80-100 lp/mm, or more, are now possible with quality small-format lenses. However, even if all aberrations are completely eliminated, which is impossible, imaging errors due to diffraction will always remain. Diffraction limits the resolution of all lenses, and the best lens, theoretically possible, is a diffraction-limited lens.

Optical diffraction is a phenomenon associated with the bending of light waves when they interact with nearby obstacles in their path. Diffraction causes a beam of light to bend slightly, and spread out, as a result of passing through a narrow aperture, and the diffracted light forms a specific pattern. The metal blades of a circular lens aperture, for example, form a circular diffraction pattern, as seen in fig.9.

The English astronomer, Sir George Biddell Airy, first described this pattern in the 1830s, and since then, it is referred to as the Airy diffraction pattern. It presents itself as a bright central disc, the Airy disc, which is surrounded by a set of concentric rings of ever decreasing brightness. The diameter (dairy) of the Airy disc is given by:

where ‘λ’ is the wavelength of light, ‘v’ is the distance from lens to image, and ‘d’ is the diameter of the circular lens aperture (see fig.5). If the lens is focused at infinity, the calculations for the diameter and radius of the Airy disc simplify to:

d_airy = 2.44·λ·N     r_airy = 1.22·λ·N

where ‘λ’ is the wavelength of light, and ‘N’ is the lens aperture in f/stops. The Airy disc receives approximately 84% of the diffraction pattern’s light energy, where the subsequent rings only receive 7, 3, 1.5 and 1%.

Optical diffraction affects the behavior of all light, including the single beam of a point light source. This means that a single image point cannot be smaller than its relevant diffraction pattern. Or, in more practical terms, the smallest possible image point is of the same size as the Airy disc. This fundamentally limits the resolution of any optical system. When observing double stars through a telescope in the 1870s, the English physicist John William Strutt (3rd Baron of Rayleigh) discovered that two stars could just be resolved if their diffraction patterns were at least as far apart as the radius of the Airy disc (fig.10). Since then, this limiting relationship between diffraction and resolution is known as the Rayleigh criterion.

fig.9       Diffraction causes a beam of light to slightly bend and spread out as a result of passing through a narrow aperture, while forming a circular diffraction pattern.

Strictly speaking, a distinction has to be made between point resolution and line resolution, because the human eye responds differently to points and lines. However, the Rayleigh criterion refers only to an approximate relationship between diffraction and resolution, and empirical data shows that it works well for photographic purposes, where minute detail has a variety of shapes.

fig.10a-d A single image point cannot be smaller than its relevant diffraction pattern. This fundamentally limits the resolution of any optical system. The Rayleigh criterion states that two image points can only be resolved if their diffraction patterns are at least as far apart as the radius of the Airy disc.

The minimum negative resolution (Rmin), necessary to achieve the required maximum circle of confusion for each negative format (see fig.4), is given by:

where ‘creq’ is the required circle of confusion for either standard or critical observation. Diffraction-limited systems achieve the highest possible lens resolution, because, according to the Rayleigh criterion, they are only limited by the radius of the Airy disc. Maximum lens resolution (Rmax) is given by:

where ‘rairy’ is the radius of the Airy disc, ‘λ’ is the wavelength of light, and ‘N’ is the lens aperture in f/stops. Fig.11 shows the diffraction limits for three wavelengths at 650 nm (infrared), 555 nm (the human eye’s sensitivity peak) and 400 nm (ultraviolet). Diffraction increases, while aberrations are reduced, as the lens is stopped down. At f/11 or above, lens aberrations are significantly reduced, but diffraction starts to seriously inhibit lens resolution.

fig.11     The actual negative resolution is limited by lens aberrations and diffraction. When a wide-open lens is stopped down, negative resolution increases at first, because lens aberrations are reduced. Negative resolution then peaks at an ‘optimal’ aperture for that lens. Stopping the lens down further decreases negative resolution again, due to continuously increasing diffraction. At very small apertures, diffraction is the only limiting factor of negative resolution.

From fig.11, we see that the digital, 16x24mm DX format barely satisfies standard observation requirements, even under the best of circumstances. Critical observation requirements are hopelessly out of reach. The 35mm format fully satisfies standard observation requirements but cannot yield a print ‘resolved beyond human detection’ either. However, stopping down to about f/8-11 provides maximum lens performance and satisfying prints. A negative made with a high-quality medium-format lens at f/8-11 can be enlarged to a print that stands up to the most critical observation, with little room to spare, but it should not be stopped down beyond f/16 to avoid diffraction. A 4x5 lens performs best at about f/11, but if required, it can be stopped down to f/32 and still achieve the critical resolution necessary for a highly detailed print.

Given a shake-free exposure, many medium- or large-format lens and aperture combinations yield a negative resolution high enough to satisfy even the most critical observer. Nevertheless, take a close look at ‘Sharpness in the Darkroom’ to make sure you transfer this detail from negative to print. Also, seriously consider the image-quality limits of diffraction when stopping down a lens, because localized softness of secondary image areas is often far less critical than uniform, but mediocre, front-to-back image detail.

The actual lens-resolution values in fig.11 are based on my equipment, materials and procedures. To determine the capabilities of your system, prepare a set of negatives depicting the USAF/1951 test pattern in fig.1a at various lens aperture settings. Subsequently, determine your negative resolutions according to fig.1c, and use fig.4 to compare the results with the negative resolution required to support standard or critical print observation.

As soon as the radius of the Airy disc is larger then the required circle of confusion, the optical system is limited by diffraction. It is, therefore, futile to compute the depth of field using a circle of confusion smaller than the radius of the Airy disc. As a consequence, the smallest circle of confusion (cmin) that needs to be taken into account is given by:

c_min = r_airy = 1.22·λ·N

where ‘rairy’ is the radius of the Airy disc, ‘λ’ is the wavelength of light, and ‘N’ is the lens aperture in f/stops. We cannot improve image quality beyond the quality limits of the entire system. Image quality is ultimately limited by diffraction.

The table in fig.12 lists the diffraction limits in the form of the maximum possible resolutions and the smallest necessary circles of confusion, depending on the lens aperture selected. Like fig.11, the table shows that the potential resolution values for f/4 to f/8 challenge the best of lenses, while even mediocre lenses have no trouble delivering the diffraction-limited resolutions of f/32 to f/90. Fig.12 also indicates diffraction-limited aperture settings for the most popular negative formats. Stopping the lens down further creates a diffraction-limited circle of confusion, which is larger than the one permitted by critical viewing (see fig.4). In other words, stopping the lens down beyond these limits prevents achieving the minimum negative resolutions required for critical viewing. In these cases, either open the aperture, if possible, or do not consider these negatives for critical viewing.

fig.12     There are diffraction-limited aperture settings for all popular negative formats. Stopping the lens down beyond these limits will prevent achieving the minimum negative resolutions required for critical viewing. Note that neither the digital DX nor the small 35mm format is suitable for critical viewing conditions, because they cannot realistically obtain the minimum resolutions necessary.

Note that neither the digital DX nor the small 35mm format are suitable for making prints that must conform with the stringent requirements for critical observation. Their lenses, film or camera sensors cannot deliver the minimum resolutions necessary to comply with this high quality standard.

Sharpness and Image Clarity

Creating sharp images is a popular topic of photography, but when photographers start talking about sharpness, they quickly struggle to find precise terminology. This is because they are referring to the visual perception of clear image detail in a photograph, and as with all human impressions, perceptions can be felt but not measured. Detection, resolution, acutance and contrast, however, are aspects of image clarity, and they can be measured. It’s similar to ‘temperature’ and ‘heat’. One is a measurable phenomenon; the other vaguely describes our human perception of it. As a consequence, we define sharpness as the visual perception of image clarity, and in general conversation, we can safely assume that ‘perceived’ sharpness always refers to a mixture of resolution, acutance and contrast.

Resolution, acutance and contrast are inseparably linked to each other, but they are based on different fundamental principles. Resolution is defined as the ability to record distinguishable fine detail. A lens that records more line pairs per millimeter than another offers more resolution. Acutance, on the other hand, is defined as edge contrast, which is the ability to clearly record finite edges between adjacent elements. A black line on a white background is perceived as perfectly sharp if the change from white to black is abrupt (high edge contrast). The smoother the transition from white to black is, the less sharp the line appears. In other words, the higher the edge contrast, the higher the acutance and the sharper the edge. Finally, contrast is a measure of differences in brightness between tones in a photograph. It’s the difference between light and shadow, and without that difference, there is nothing to see. There is full contrast between black and white lines, but little or no contrast between gray lines. The more contrast there is between lines, the easier they are to see, and thus, the sharper they appear.

fig.13     Increasingly sharper-appearing lines, with their respective density traces below each, illustrate how perceived sharpness increases with edge contrast.

fig.14     Resolution, acutance and contrast are very different measures of image clarity, and sharpness depends on the complex interaction between all three.

Fig.13 explores different degrees of acutance and illustrates how perceived sharpness increases with edge contrast. Shown are four increasingly sharper-appearing lines, and below each is a density trace across the respective line. The density trace of line (a) has a very smooth density transition from the light-gray background to the dark-gray line. This line does not appear to be sharp at all. Instead, it seems to be totally out of focus and rather blurry. The density trace across the next line (b) shows a more abrupt change in edge density, but the increases and decreases still follow a fairly smooth density transition, which also makes for the appearance of a slightly out-of-focus, unsharp line. The next line (c) is optimally sharp, featuring harsh, clearly defined edges in the density trace. In practice, it is not possible to achieve this high level of acutance with standard pictorial film and full tonal development, but with quality optics, special high-contrast copy films can deliver acutance this high. Nevertheless, it is possible to artificially increase the acutance and get an even sharper line than line ‘c’, by utilizing the concept of increased edge contrast to its fullest. This can be done in both analog photography and digital imaging. In analog photography, we have a choice between special acutance film developers and unsharp masking, which is discussed in its own chapter. Both methods achieve a line and a density trace similar to the example shown in fig.13d. In digital imaging, almost identical results are obtained, because sharpening algorithms mimic the principle of exaggerated acutance, although with an unfortunate tendency to overdo it.

Fig.14 highlights the complex interaction between resolution, acutance and contrast. It shows the same line pattern with increasing resolution from left to right and decreasing acutance and contrast from top to bottom. Pattern ‘a’ has optimal sharpness due to the high edge contrast of each line and the full contrast between each line. In pattern ‘b’, the lines are not as clearly defined, because the edge contrast is reduced (black lines start to ‘bleed’ into white lines), and consequently, the contrast between lines is slightly reduced. Pattern ‘c’ seems even less sharp with very smooth line transitions, which result in low contrast between lines. Towards the high-resolution end, the lines actually blend together completely. At some point, no line pattern can be resolved, because there is no contrast left between lines. Patterns ‘d’ and ‘e’ are similar to ‘c’ but the initial, full pattern contrast is reduced to 50% and 10%, respectively, which decreases image clarity even further. As we can see from this example, resolution, acutance and contrast are very different measures of image clarity.

fig.15     Test patterns are useful when exploring technical issues, but we get a better understanding for how the aspects of sharpness influence our photography when we study their impact on our real-life images.

(image © 2008 by Artlight Studios, all rights reserved)

Test patterns are useful when exploring technical issues, but we get a much better understanding for how the aspects of sharpness influence our photography, when we study their impact on our real-life images. Fig.15 shows an example of how resolution, acutance and contrast influence image clarity. In fig.15a, an attempt is made to compensate for low image resolution with increased acutance and contrast. At close inspection (where resolution counts the most), the attempt fails, but at arm’s length, the image appears to be sharper than the next. fig.15b is of high resolution, but a low edge contrast keeps image clarity below expectations. In fig.15c, high resolution is successfully supported by high acutance, and in conjunction, they make for the sharpest image of the three. These examples show that increased acutance and contrast may be able to overcome a limited lack of resolution. But, we can safely conclude that a truly sharp image depends on high resolution, acutance and contrast.

Modulation Transfer Function (MTF)

Apart from camera sensors or film, lenses are indubitably the most important contributors to image quality. This does not mean that being the proud owner of a good lens is a guarantee for creating good images, but it does provide a solid foundation, and without a sharp lens, it’s impossible to get a sharp image. This probably explains why so many photographers feel the need to test their lenses right after the purchase, or why they spend so much time and energy to acquire, and inquisitively study, published lens tests before they invest in a new lens.

A practical and convenient way to measure the optical quality of a lens, without specialized laboratory equipment, is to take a photograph of a resolving power chart such as the USAF/1951 test pattern. As previously explained, resolution measurements alone are not representative of image clarity, but they are a reasonably reliable measure of the fundamental recording characteristics of a lens.

The lens-resolution limit is determined by inspecting the negative with a loupe and finding the smallest, still resolved, line pattern (see fig.1). The benefit of this test method is its simplicity, and if the test is conducted with the photographer’s favorite camera, tripod, film, developer and so forth, it’s also a reasonable system test. However, because perception and judgment are involved, the test results are highly subjective. The element resolution of the USAF/1951 test pattern increments in 12% steps, and observers rarely agree on the same element representing the highest resolution (fig.16). Also, there is an optimum viewing distance or magnification. If the magnification is too low, the eye cannot separate the smallest, still resolved, line pattern. And, if the magnification is too high, an otherwise resolved line pattern is lost in the noise of micro detail and is not recognized as a coherent pattern, which is why a high-magnification microscope would be of no use. All this makes a 12% variance in test results likely and a 25% variance possible. Nevertheless, a disciplined practitioner, working with reasonable care and consistency, will find this to be a valuable and practical method for comparative testing.

The introduction of the modulation transfer function (MTF) addressed many shortcomings of simply photographing an ordinary line pattern. Today, MTF is the standard scientific test method to evaluate optical lens quality, and simple resolution tests have fallen from favor. Conducting an MTF test is typically beyond the means of an amateur photographer, but it’s still worthwhile being able to read and understand MTF charts, because a major benefit of these charts is that they illustrate the complex interaction between resolution, acutance and contrast, which we perceive as sharpness. MTF charts have better correlation to lens quality than resolution measurements alone.

fig.16     A disciplined practitioner, working with reasonable care and consistency, will find photographing test patterns to be a valuable and practical method for comparative testing, but the test results are subjective.

In brief, MTF is the spacial frequency response of an imaging system or component. It is the optical equivalent of acoustic frequency response plots commonly produced for audio systems. The difference is that for audio systems, the frequency is measured in cycles per second (Hz), and for optical systems the frequency is measured in cycles per millimeter (cycles/mm). In both cases, however, the response is measured as a function of the input frequency. This sounds a lot more difficult than it actually is, because the essential principle of the MTF is rather simple (fig.17). Take a well-defined input pattern (a), photograph it (b), and compare the output pattern to the input pattern (c). The ratio of output versus input contrast is called modulation transfer factor, and measured for numerous spacial frequencies (d), results in the MTF.

fig.17a-d The essential principle of the modulation transfer function (MTF) is rather simple. Take a well-defined input pattern (a), photograph it (b), and compare the output pattern to the input pattern (c). The ratio of output versus input contrast is called modulation transfer factor, and measured for numerous spacial frequencies (d), results into the MTF, which is a sophisticated and objective optical performance measure of lens quality.

The optimal test target for an MTF evaluation is a sinusoidal pattern, consisting of progressively thinning black and white lines (increasing frequency), whose densities blend smoothly into each other (fig.17a top). A density trace across such a pattern is a sine wave of increasing spacial frequency but consistent amplitude and, consequently, consistent contrast (fig.17a bottom). When such a test pattern is photographed and compared to the original pattern from left to right, low-frequency line patterns on the left are almost identical to the original, but high-frequency patterns on the right are not as clearly recorded (fig.17b top). If the spacial frequency is high enough, the lines eventually merge and blend into a medium gray, leaving no contrast or distinguishable line pattern at all. A density measurement across the pattern from left to right shows that the black line peaks are getting progressively lighter and the white line peaks are getting progressively darker. While the spacial frequency increases, the contrast between black and white lines diminishes, and eventually, there is no contrast left. The pattern disappears into a medium gray. A density trace across the output pattern illustrates this through a continuous loss of amplitude, ultimately leveling out at zero contrast (fig.17b bottom).

The measurement examples in fig.17c show a contrast reduction for spacial frequencies of 10, 20 and 40 cycles/mm, to 95, 80 and 20%, respectively. The ratio of output versus input contrast is called the modulation transfer factor. In practice, the transfer factors of numerous spacial frequencies are calculated, using a multitude of micro-densitometer measurements. After the data is collected, the modulation transfer factors (vertical axis) are plotted against their respective spacial frequencies (horizontal axis), forming the modulation transfer function, as shown in fig.17d.

A contrast response of above 80% is considered to be a good contrast performance, and 50% is still acceptably sharp, but at 10% the image contrast is so severely attenuated that this is considered to be the limit of optical resolution, regardless of the fact that under favorable viewing conditions, contrast responses down to 1% still allow for a line pattern to be perceived. Nevertheless, 10% image contrast roughly corresponds to the Rayleigh criterion, which is generally accepted as the practical resolution limit.

Fig.18 shows a line pattern photographed with three different lenses and compares them with their respective MTFs. Lens ‘a’ represents an unrealistically perfect lens. The recorded image is identical to the test target, and because of this, the MTF is a horizontal line with a 100% modulation transfer factor. This is the ultimate optical performance. Lens ‘b’ is a high-contrast, high-acutance lens of limited resolution, and lens ‘c’ is a low-contrast, low-acutance lens with high resolution. High contrast and acutance do not necessarily mean high resolution. A lens delivering both high contrast and resolution is an optical design challenge. As seen in the patterns of lines ‘b’ and ‘c’, the high-contrast lens ‘b’ appears to be sharper and more brilliant than the high-resolution lens ’c’. When it comes to perceived sharpness, contrast and acutance are often more important than resolution.

It’s worth noting that MTF tests are often conducted with sinusoidal test targets, as well as line targets. Strictly speaking, spacial frequencies of sinusoidal patterns are measured in cycles/mm, and the resolution of a line pattern is measured in lp/mm. Comparing them directly is not entirely correct, but the test results only show small differences with no practical consequence, and therefore, both units are commonly used interchangeably.

Simple MTFs, such as the ones shown in fig.17-18, are typically prepared for a variety of optical components and systems. You’ll find them for films, paper, scanners, camera sensors and other light-sensitive materials, including the human eye! When it comes to lenses, they don’t tell the whole story, because lenses project the light into an image circle, but the negative format crops this image circle to the familiar square or rectangular shape. Lens quality is best at the image center and gradually worsens towards the edge of the image circle. To more realistically represent lens quality, lens MTFs are limited to a few spacial frequencies, but show the modulation transfer factor across the entire negative format (see fig.20).

Fig.19 shows how lens MTFs are prepared. Small test targets, with fixed spacial frequencies, are placed at a strategic location in the image area. This is done from the center towards the edge of the image circle and up to the corner of the negative format. The test targets include two sets of test patterns, one tangential and one sagittal (radial) to the image circle, because lens performance is not uniform in both directions. Once all test data is compiled, typical lens MTFs can be prepared. Fig.20 shows three medium-format examples, one for a wide-angle, one for a normal and one for a telephoto lens. In typical lens MTFs, the modulation transfer factors (vertical axis) are plotted against the distance from the image center (horizontal axis). Each graph shows the tangential and sagittal lens performance at 10, 20 and 40 lp/mm for one particular focal length and aperture.

A detailed lens evaluation can be conducted from these graphs, if we consider and accept the different spacial frequencies as being representative of different lens performance criteria. The 10-lp/mm line is a good indicator for the contrast behavior of the lens. The 20-lp/mm line represents ‘perceived’ sharpness, and the 40-lp/mm line illustrates the lenses’ resolution limits across the negative format. In general, the higher the transfer factors and the straighter the lines are, the better the respective lens performance is. However, what follows are some commonly agreed guidelines, which support a more detailed analysis of lens MTF charts.

fig.18     The photographs of a line pattern, made with three different lenses, are compared and correlated to their respective MTFs. Lens ‘a’ represents an unrealistically perfect lens. Lens ‘b’ offers more contrast but less resolution than lens ‘c’. High contrast and acutance do not necessarily mean high resolution. The high-contrast lens ‘b’ appears to be sharper and more brilliant than the high-resolution lens ’c’. When it comes to perceived sharpness, contrast and acutance are often more important than resolution.

fig.19     A medium-format lens MTF is prepared by placing small test targets, with fixed spacial frequencies, at strategic locations of the image area. This is done from the center towards the edge of the image circle and up to the corner of the negative format.

Lenses with 10-lp/mm transfer factors (low frequency) of 90% or better have excellent contrast. For a lens to be perceived as truly sharp, 20-lp/mm transfer factors must be around 80% at the image center and not drop below 50% at the borders. A lens is considered to have good resolution if it has 40-lp/mm transfer factors (high frequency) of above 60% at the center and not less than 20% at the image borders.

fig.20a-c In these lens MTFs, modulation transfer factors are plotted against the distance from the image center to create the modulation transfer function. Each graph shows the tangential and sagittal lens performance at 10, 20 and 40 lp/mm for one particular focal length at a typical working aperture.

In general, but not always, longer focal-length lenses are superior to wide-angle lenses, especially at the image corners. When comparing lens performance, only lenses of the same or similar focal length should be judged. The same is true for lens apertures. Wide-open and fully stopped-down lenses don’t perform as well as lenses that are stopped down a stop or two to a more realistic ‘working’ aperture. Never compare a wide-open MTF of one lens to a working-aperture MTF of another. And, don’t be overly concerned with the lens performance on the very right-hand side of the MTF chart. Much of it is dedicated to the small corner areas of the negative format. For this reason, these areas are grayed in fig.20a. But, some attention should be given to large performance variances between the tangential and sagittal lines. This indicates the presence of a lens aberration called astigmatism, which, among other things, results in a poor ‘bokeh’. Bokeh is a Japanese word, describing the way in which a lens reproduces out-of-focus images.

Despite the complexity of generating them, and the learning curve required to read them, lens MTFs are a valuable method to evaluate absolute lens performance. We need to be aware, however, that some lens manufacturers generate their MTFs from lens-design computer models and not from actual test data. This is better, of course, than having to live with the choice of some lens manufacturers not to generate or publish their MTFs at all. Due to lack of a standard, you may not always find lens MTFs prepared for the same spacial frequencies. Large-format lens MTFs, for example, are often produced for 5, 10 and 20 lp/mm.

MTFs have some inherent limitations. They don’t tell us anything about most lens distortions or vignetting, and lens MTFs don’t give us a numerical value for the highest resolution obtainable. But, with an MTF at hand, and combined with our own comparative testing, we have all we need to understand the important performance characteristics of our lenses, including sharpness.

 

Critical Focusing

What you see is what you get?

Prior to picture taking, we typically focus the image on a view screen, and during the actual exposure, the image is projected onto the film plane. While doing so, we take for granted that view screen and film plane, despite residing at two different locations, have the same distance from the lens.

Camera manufacturing is about balancing process capabilities with customer expectations to achieve a required mechanical accuracy within acceptable tolerances. In addition, all mechanical devices are subject to unavoidable wear and tear, which require periodic adjustment or replacement. To manufacture within tolerance is no guarantee that the product will stay that way forever. Within twelve months, we once had to adjust a professional medium format SLR, two medium-format rangefinders and a well-known make of 35mm rangefinder. One of these cameras was brand-new. After being adjusted, they all focus perfectly, putting the initial camera setup in question, and proving that the following test method is valid.

fig.1       The grid of an enlarging easel or a cutting board makes a perfect focus target for checking critical camera focus.

What Is Reasonable?

Take, for example, a 90mm, f/2 lens on a 35mm rangefinder. Clearly, the f/2 aperture is not for viewing brightness, but is designed for picture taking. The tolerances of the camera body, lens and photographer add up. The human element in any focus mechanism provides opportunity for error, but it is not an unreasonable assumption that the mechanical focus accuracy should be within the depth of field at the maximum lens aperture. With the 90mm lens at the minimum focus distance, the acceptable depth of field is 10 mm at most. For a portrait, this is the difference between acceptable and unacceptable eye sharpness. The alignment between view screen and film plane must be well within the depth of focus, which in this example, is a tight tolerance of less than ±0.05 mm.

A Simple Focus Target

For any kind of focus check, we need to be able to set up the camera with perfect repeatability. A good focus target must be easy to focus on and, at the same time, indicate the magnitude of error in focus. This suggests a series of horizontal markings along the optical axis. However, since most split-image and rangefinder screens are better at determining vertical than horizontal lines, adding a series of vertical lines makes good sense. Put these together and you get a grid.

Rather than drawing a unique grid, we can use a piece of graph paper, our cutting-mat scale or the grid on our enlarger easel, all of which make adequate focus targets. For this example (fig.1), we use the grid on an enlarging easel, which is a white piece of plastic with fine, black grid lines in 20mm increments.

The camera is set up on a tripod and carefully focused on the 100mm mark, using the vertical lines for critical adjustment. Additionally, the camera is at an angle of about 30° to the easel plane and close to the minimum focus distance. One benefit of focusing rangefinder cameras is immediately apparent when viewing the grid. Since the rangefinder and viewfinder window have a different perspective on the grid, the vertical grid lines have different slants and seem to cross over at the point of focus. Consequently, this enables extremely accurate focus adjustment. With split-image viewfinders, position the split line on the focus point.

As can be seen in fig.1, the gradual blurring of the vertical lines clearly identifies the focus point along the scale, aiding accurate focus measurement. At the same time, it is possible to estimate the range of useful focus at this short range.

We suggest that you repeat the test a few times to ensure your technique. With rangefinder cameras, try arriving at perfect focus from near and far distance settings, to check for any play in the mechanism.

fig.2       Typical film thickness and ANSI film-holder dimensions in inches

Improving View Camera Focus

When using a view camera, the image is composed and focused on the ground glass. One surface of the ground glass is textured to provide a means for focusing the image. It is important that this textured surface faces the lens, because it is the image forming side. To take an exposure, the ground glass is replaced by the film holder. At this point, the film must be in the same plane as the ground glass was during focusing, so the negative is perfectly sharp. Camera backs and film holders are machined to tight tolerances to ensure this condition (fig.2).

A well-focused image and full utilization of the intended depth of field are achieved if these tolerances are close to zero. Small deviations can be tolerated, because the depth of focus for view cameras is relatively large (1 mm or 0.040 inch for a 4x5 negative at f/5.6), but even small tolerances will shift the focus and depth of field. It is, therefore, important to keep the ground glass in perfect alignment with the film plane. Fig.2 shows typical film thickness and the ANSI standard dimensions for film holders in inches. However, experience shows that many cameras and film holders deviate enough from these standards to warrant a simple check.

fig.3       A steel ruler, a toothpick and a paper clamp are used to measure the location of the film plane in a 4x5-inch sheet-film holder in relation to the open camera back.

fig.4       The same setup is used to check for a proper ground-glass location after the film holder is removed, and the toothpick is clamped to an average film-holder depth.

The previously discussed focus target works well for SLRs and rangefinder cameras, but is not ideal for view cameras, the reason being that each test exposure checks only one side of one film holder. It is not uncommon to have a dozen film holders or more, and making dozens of test exposures is time consuming and costly.

A Simple Check

In his May/June 1999 Photo Techniques magazine article, Jack East Jr. proposed a simple but effective alternate method to check whether the ground glass and the film plane are within acceptable tolerance.

Place a piece of film into a holder and insert it into the camera back. Remove the back from the camera, and lay it flat on a table as shown in fig.3. Rest the edge of a rigid ruler across the camera back. Hold a toothpick or cocktail stick vertically against the ruler, lower it until it touches the film and clamp or tape it to the ruler, thereby identifying the film plane location. After doing this with all film holders, leave the toothpick positioned for an average holder.

fig.5a     A Fresnel lens can be added to an existing camera back simply by placing it behind the ground glass, in which case, the ground glass maintains its alignment with existing film holders. However, image formation on two separate surfaces can make accurate focusing difficult.

fig.5b     The Fresnel lens can be added in front of the ground glass as well, so image formation takes place on only one surface. However, the ground glass is no longer aligned with the film plane, and the camera back must be machined or otherwise adjusted to regain proper focus.

Now, remove any film holder from the camera back, and compare the average film plane with the ground glass location (see fig.4). If the toothpick just touches the ground glass, then no adjustments are required. Knowing that a sheet of regular writing paper is about 0.1 mm (0.004 inch) thick provides a convenient measuring device to quantify any offsets. If the toothpick touches before the ruler, then you can shim the ground glass with paper. If there is an unacceptably large gap between toothpick and ground glass, then professional machining of the camera back is required.

With the toothpick still positioned to identify the average film plane location, measure all film holders for variation. According to the standard in fig.2, a tolerance of ±0.007 inch, or two layers of paper, is acceptable for the 4x5 format. Discard or avoid film holders outside this tolerance.

Using a Fresnel Lens

One variation in ground glass design is the addition of a Fresnel lens. Its purpose is to provide even illumination over the entire ground glass, making focusing, especially in image corners, significantly easier. A Fresnel lens is typically a flat piece of plastic, with one side built up from a series of thin concentric rings, which function like a lens. The rings are usually barely perceptible to the naked eye, but become obvious when viewed through a focus loupe.

A Fresnel lens equalizes image brightness when placed either in front of or behind the ground glass, and there are some pros and cons with each setup. When a Fresnel lens is added to an existing camera back, it is far simpler to place it behind the ground glass as shown in fig.5a. The ground glass retains its position, and the alignment with existing film holders is maintained. However, in addition to image formation on the textured surface of the ground glass, it is possible to focus an image on the ridges of the concentric rings of the Fresnel lens. The image formation on two separate surfaces can make accurate focusing difficult, but with practice, this is rarely an issue.

Alternatively, the Fresnel lens can be added in front of the ground glass as seen in fig.5b. This has the advantage of image formation only taking place on one surface, since the ridges are in contact with the textured surface of the ground glass. However, if the Fresnel lens is added to an existing camera back, the disadvantage is that the ground glass, and the associated focus plane, is out of its original position. Consequently, the focus plane is no longer aligned with the film plane, and the camera back must be machined or adjusted to allow for the Fresnel lens thickness. In either setup, make sure that the textured surface of the ground glass faces the lens and is aligned with the film plane, and that the ridges of the Fresnel lens are facing the ground glass.

fig.6       An advanced focus target provides quantifiable results.

An Advanced Focus Target

A simple focus target, such as the grid on our enlarger easel in fig.1, is more than adequate to verify camera focus once in a while. But, if you intend to conduct a lot of focus testing, or you need quantifiable results, you might want to invest the time in building a more sophisticated focus target. As an example, our advanced focus target in fig.6 provides repeatable and quantifiable results and is easily made within an hour.

As shown in fig.6, take some mat-board scraps and construct a 45° triangle from it. Make it about 25 mm thick and 150 mm tall. Then, copy the focus scale in fig.8 and glue it to the long side of the triangle. The focus scale is elongated along the vertical axis to be at the correct dimensions if viewed foreshortened under 45°. Building the surrounding support is an option, which makes repeatable focusing a lot easier. When using a support, make sure the focus planes of the support structure line up with the zero marking on the focus scale, before you level the camera and take the picture with a wide-open aperture.

Fig.7 shows two sample test images. The image on the left shows a far-sighted focusing error of about 5.5 mm, prior to the camera adjustment. The image on the right verifies perfect focus after such adjustment.

fig.7       These test images were taken from a distance of 935 mm at f/1.8 with an 85mm lens (m=0.1). The image on the left shows a far-sighted focusing error of about 5.5 mm (0.6%), prior to camera adjustment. The image on the right verifies perfect focus after adjustment.

fig.8       This is our advanced focus scale at full size. It is already elongated along the vertical axis to be at the right dimensions if viewed foreshortened under 45°.

A Practical Hint

Focusing a camera in low-light situations is not an easy task. We would like to share a proven technique, which works well even in the darkest church interiors.

Purchase two small flashlights for your camera bag. Mag Instrument is a popular brand, which comes in many sizes. Unscrew the tops, which turns them into miniature torches, and place them upright into the scene at the two extremes of the desired depth of field (fig.9). Focusing on the bright, bare bulbs is simple, no matter how dark the location is.

fig.9       Focusing on the bright bulbs of miniature flashlights is simple, no matter how dark the location is.

 

Pinhole Photography

The fascinating world of lensless imaging

A number of dedicated individuals paved the way for the invention of photography with their accomplishments in several areas of the natural sciences. However, in very basic terms, photography requires only one condition to be satisfied, the successful combination of image formation and image capture.

Image capture has been in the chemical domain for over 150 years, but modern electronics recently added digital image capture as a realistic alternative and provided us with fresh tools for image manipulation. Image formation, on the other hand, was always governed by the laws of optics. It may be of historic interest to note that image formation and capture were practiced independently for some time, before they were successfully combined to make photography possible. Nevertheless, taking a closer look at these building blocks of photography, one quickly finds that image formation is far older than image capture.

fig.1       This is thought to be the first published picture of a camera obscura and a pinhole image, observing the solar eclipse of 1544-Jan-24, in the book De Radio Astronomica et Geometrica of 1545 by Gemma Frisius.

© 2001 by Andreas Emmel, all rights reserved

fig.2       A print made with an 11x14-inch large-format pinhole camera shows surprising detail and clarity.

fig.3a     Simply holding up a card in front of a subject is not sufficient to create an image, because every point on the card receives light rays from numerous points on the subject.

fig.3b     But if an opaque panel, containing a tiny pin-sized hole, is placed between the subject and the card, the panel blocks all light rays coming from the subject with the exception of a limited number entering through the pinhole. The small hole restricts the light rays coming from the subject to a confined region, forming countless blurry image circles and a fuzzy image.

fig.3c     To improve image quality, the pinhole is replaced by a lens. It converges several light rays from the same subject point into one focused image point. This makes for a sharper and brighter image than a pinhole can possibly provide.

Basic image formation is as old as nature itself. The simplest arrangement for basic image formation is by way of a pinhole. The overlapping leaves in trees form numerous pinholes naturally, through which countless sun images are projected onto the ground. It is conceivable that humans were captivated by the crescent pinhole images of an eclipsed sun as early as the dawn of mankind.

The earliest known description of pinhole optics came from Mo Ti in China from around 400 BC, and Aristotle wrote about his observations of the formation of pinhole images in 330 BC. The first known proposals to create a small opening in an otherwise darkened room (camera obscura), in order to intentionally produce pinhole images, came from Alhazen in Egypt around 1020 AD and Roger Bacon (1219-1292) in England. Obsessed with representing realistic perspectives, Renaissance artists, including Leonardo da Vinci (1452-1519), often used a camera obscura to develop the early sketches for their magnificent paintings. In 1584, the second edition of Giovanni Battista Della Porta’s book Magia Naturalis was published. In this book, he describes the formation of pinhole images and the construction of a pinhole camera in detail. Around that time, Johannes Kepler (1571-1630) coined the phrase camera obscura, which literally means ‘dark room’. Soon after, many pinholes were replaced by a simple concave lens, which improved image brightness and quality. Pinhole imaging languished over 200 years, until after the invention of photography, to have its first revival around 1850.

Image Formation

Image formation starts with light rays, which are either emitted or reflected by the subject. The light falling onto an opaque subject is partially absorbed and partially reflected. Theoretically, reflection is either directional (specular) or multidirectional (diffuse). In reality, the actual reflection depends on the surface characteristics of the subject and is always a mixture of specular and diffuse reflections. Smooth surfaces, such as glass, mirrors, polished metal or the calm surface of a lake, create predominantly specular reflections. Rough surfaces, such as leaves, stone, cloth or dry skin, create primarily diffuse reflections.

For the purpose of investigating general image formation, we can safely assume that every point of an illuminated subject emits or reflects light in multiple directions. Simply holding up a card in front of the subject is not sufficient to create an image on the card, because every point on the card receives light rays from numerous points on the subject (see fig.3a). Successful image formation requires a more structured approach of correlating subject with image points.

The simplest arrangement for image formation is achieved by placing a flat opaque object, containing a tiny pin-sized hole, between the subject and the card (see fig.3b). The opaque panel blocks all light rays coming from the subject with the exception of the few entering through the pinhole. The hole is small enough to restrict the image points on the card to light rays coming from a confined region of the subject, forming countless blurry image circles, which together form a dim fuzzy image. This way, compromised image formation is possible, because every potential image point receives light rays only from a limited number of subject points.

As we can see, expensive optics are not essential to the image-forming process, but to improve image quality beyond the pinhole, the light-restricting opening must be replaced by a convex lens. The lens converges several light rays from the same subject point into one focused image point through refraction (see fig.3c). This makes for a sharper and brighter image than a pinhole can possibly provide. High-quality image formation is only possible with a lens, where every potential image point receives light rays exclusively from its corresponding subject point. Nevertheless, pinhole photography offers a subtle beauty, which is difficult to achieve otherwise and, therefore, makes exploration and optimization of this fascinating field of photography worthwhile.

Making Your Own Pinhole Camera

The first step in building a pinhole camera is to create the pinhole itself. A high-quality pinhole is accurate in diameter and has a smooth perimeter for superior image clarity. The smoother the edge of the pinhole is, the sharper the resulting pinhole image will be. You can buy a pinhole or make one yourself.

Several suppliers of optical and scientific products sell laser-cut pinholes, which are typically drilled into thin brass foil. Professionally made, laser-cut pinholes do not cost a lot, which makes them the best choice, because they are also extremely precise in diameter and have an exceptionally smooth edge (fig.4b). Nevertheless, if you are in a rush, or just want to experiment with a pinhole, you can simply take a pushpin or sewing needle and force it through a piece of black cardboard (fig.4a). This will make for a workable pinhole, but don’t expect an optical miracle, because the rough edge will degrade image quality significantly.

fig.4a     Simply forcing a needle through a piece of cardboard will result in a workable pinhole, but the rough edge degrades image clarity.

fig.4b     A laser-cut pinhole, with a particularly smooth perimeter, gives the best possible image quality.

If you aim for more accuracy, consider the following work instructions, illustrated in fig.5. This will not provide you with a pinhole of ultimate precision, but with a bit of practice and the right materials, a good-quality pinhole can be made within a few minutes.

1.   Use scissors to cut a piece of metal from brass foil, or an aluminum can, roughly 15x15 mm in size.

2.   Place the metal flat onto a soft wood support, and firmly press a ballpoint pen into the center of the square, creating a clearly visible indentation.

3.   Turn the metal over, and use fine sandpaper to thin away the bump without penetrating the metal.

4.   Create the pinhole by pushing a needle through the center of the indentation, and gently reinsert the needle from the other side to smooth the edge.

fig.5a     With a little bit of practice and the right materials, a good-quality pinhole can be made in a few minutes.

fig.5b     The pinhole material thickness limits the angle of coverage. Thick materials may reduce the angle of view, and the pinhole will no longer fill the entire negative format.

The pinhole material thickness is of some consequence to the pinhole image, because it limits the angle of coverage. A thickness of about 0.1 mm is ideal, because it provides an angle of over 125°. Thicker materials may reduce the angle of view, and the pinhole will no longer fill the entire negative format (see fig.5b).

fig.6       Old medium-format camera bodies make perfect pinhole cameras. This shows a well-kept 6x9 box camera from around 1930 after the conversion.

It is a good idea to measure the pinhole diameter before the pinhole is mounted to the camera body. It is difficult to measure afterwards, and without knowing the size of the aperture, we cannot accurately determine the working f/stop of the pinhole camera. Unless you have access to a microscope with measuring capability, simply magnify the pinhole by any available means. Use a slide projector, the darkroom enlarger or a scanner to perform this task. First, prepare a measurement sample, for example two lines, known to be 20 mm apart, and enlarge or scan this sample to determine the magnification factor. Finally, enlarge or scan the pinhole at the same magnification, measure the projection or the scan and calculate the actual diameter of the pinhole. The working f/stop of the pinhole (N) is given by:

where ‘d’ is the diameter of the pinhole, and ‘f’ is the focal length of the pinhole, which is the distance between the pinhole and the film plane, assuming that a pinhole camera is always focused at infinity.

Almost any container can be turned into a pinhole camera body as long as it is absolutely light tight. Popular items include cardboard or metal boxes of all sizes, as well as cylindrical storage containers for food, chemicals or rolls of film. Everything from 35mm film canisters to full-size delivery vans has been converted to portable pinhole cameras. Best suited, and far more practical, are old camera bodies. They are already designed to safely hold and transport film, and with the exception of view cameras, most of them offer some kind of viewfinder to compose the image and a shutter to control the exposure.

Fig.2 shows a pinhole image that was taken with a self-made 11x14-inch large-format view camera. It takes minimal effort to convert a view camera into a pinhole camera. Temporarily mounting a pinhole into an empty lens plate is all one has to do to finish the conversion. This small endeavor is rewarded with large negatives and pinhole images of surprising detail and clarity, because the maximum possible resolution with contact-printed pinhole images (see fig.14) approaches the resolving power of standard human vision, which is around 7 lp/mm.

fig.7       Pinhole images have an almost infinite depth of field combined with beautiful image softness. This image softness is partially caused by diffraction but also by motion blur during long exposure times, which are rather common for pinhole photography.

Medium-format box cameras offer an opportunity for a more permanent pinhole conversion. Old medium-format box cameras are available in abundance on the used-camera market and can be obtained for little money. However, be certain to hunt for a model that works with the common 120-film format. This format was introduced in 1901 by Kodak for their Brownie No.2 and is still manufactured today, because it is used in all modern medium format cameras. Fig.6 shows my medium-format pinhole camera, based on a well-kept Balda Poka, which was made in Germany around 1930. I paid less than $15 for it in an internet auction. The simple meniscus lens was removed and replaced with a 0.38mm laser-cut pinhole. This diameter is ideal for the 6x9 negative format and the 105mm focal length. The working aperture computes to f/278 or f/256 and a 1/3 stop. The shutter has two settings, 1/30 s and ‘B’. For the long exposures, which are typical for the small apertures in pinhole photography, I use the ‘B’ setting exclusively and chose to keep the shutter open by securing the release lever with a rubber band.

The simple snapshot in fig.7, which was taken with the converted medium-format camera in fig.6, illustrates the almost endless depth of field in pinhole photography. When selecting a camera body for a pinhole conversion, be aware that many old medium-format cameras have a small red window at the back. This window is part of the manual film advance system and is provided to identify the current negative frame. The 120 roll-film format has the frame numbers of all popular medium negative formats printed on the outside of the backing paper, and they can be seen through the window. To protect the film from harmful light entering through the window, it is made of red-tinted glass or plastic. This protection works well for orthochromatic films but is not a reliable safeguard for modern panchromatic films. Before you load the camera with panchromatic film, cover the red window with a piece of black tape from the outside. Whenever you need to advance the film, shade the window with one hand and carefully pull the tape aside with the other. Then, advance the film to the next frame and quickly cover the red window with the tape again.

Analog or digital small-format SLRs are easily converted to sophisticated pinhole cameras by sacrificing an opaque body cap. The distance from the camera’s lens mount flange to the film or focal plane is, therefore, an approximate measure for the focal length of the pinhole. Drill a hole into the center of the body cap, and cover it by taping an appropriate pinhole to the back (fig.8). Keep the modified cap in the camera bag for quick conversions between lens and pinhole imaging.

As with lens-based images, the quality of pinhole images increases with negative size. This may be of some consequence for images that mainly require almost endless depth of field. Nonetheless, it is important to realize that the beauty of pinhole images is largely based on their diffraction-limited performance. The inherent fuzziness makes pinhole photography perfectly suited for all those images where the subject will benefit from a little softness or romantic mystery. If pinhole images were perfectly sharp, there would be little reason to make them.

The Optimal Pinhole Diameter

Realizing that pinhole images can never be perfectly sharp has not stopped photographers from seeking to optimize the quality of pinhole images and searching for the optimal pinhole diameter (fig.8). The image clarity of lens-based photography is limited by lens aberrations and diffraction. Closing the aperture reduces lens aberrations significantly but slowly increases the degrading influence of diffraction. This improves the overall image sharpness up to a point, but with decreasing apertures, diffraction eventually becomes the only limiting factor of image clarity. Obviously, a lens-less pinhole does not suffer from lens aberrations, but the image clarity of pinhole photography is limited considerably by diffraction.

fig.8       Analog or digital SLRs are easily converted to sophisticated pinhole cameras by drilling a hole into a spare body cap and covering it with a pinhole plate.

Simple geometric optics dictate that the optimal pinhole is as small as possible, because the smaller the hole, the smaller the fuzzy image circles are (see fig.3b), and the sharper the pinhole image will be. However, this ignores the influence of diffraction, which causes the light to spread, as it passes through the narrow aperture, and increases the size of the fuzzy image circles. Diffraction optics dictate that the pinhole is as large as possible to minimize light spreading. As a consequence, the ideal pinhole diameter is as small as possible and as large as necessary.

In 1857, Prof. Joseph Petzval was apparently the first to find a mathematical equation to determine the optimal pinhole diameter. Disagreeing with his proposal, Lord Rayleigh published a competing formula in 1891, which gave a much larger diameter, as did William Abney in 1895 with yet another equation. All three attempts were based on geometric optics, but no consensus was reached among photographers as to which was the ‘true’ optimal pinhole diameter. More equations, this time based mainly on empirical studies, followed until well into the 20th century. Many equations performed well enough to find enthusiastic followers, making it even more difficult to reach consensus on one optimal pinhole diameter. In retrospect, it seems like a twist of fate that Lord Rayleigh did not consider the research on diffraction by Sir George Airy from 1830, or his own diffraction criterion, which he published almost 20 years before offering his pinhole equation. Because, with his in-depth knowledge of diffraction and photography, he held the key to finding the ideal pinhole diameter, which everyone can agree to.

fig.9       Most equations to calculate the optimal pinhole diameter (d) follow the following format:



where ‘λ’ is the wavelength of light, ‘f’ is the focal length of the pinhole, and ‘k’ is a constant value, typically between 1 and 2.

fig.10     The optimal pinhole diameter (d) to optimize image sharpness is derived from the Airy disc by:



where ‘λ’ is the wavelength of light, ‘N’ is the pinhole aperture in f/stops, and ‘f’ is the focal length of the pinhole.

Remember that diffraction optics dictate that the pinhole is as large as possible to minimize light spreading, and that geometric optics dictate that an ideal pinhole is as small as possible to optimize image clarity. Considering the Airy disc and the Rayleigh criterion leads us to two theorems for an ideal pinhole diameter and suggests that there may be more than one right answer.

1.   The smallest pinhole possible is based on the Airy disc to optimize image sharpness.

2.   The largest pinhole necessary satisfies the Rayleigh criterion to optimize image resolution.

Both equations are derived, as in the example shown in fig.10, from either the Airy disc or the Rayleigh criterion. Infinity focus is assumed for both, which in reality means that they provide a depth of field from the hyperfocal distance to infinity. In both equations, the pinhole diameter is a function of the wavelength of light and the focal length of the pinhole, but a different numerical constant is used in each formula.

fig.11     The MTF graph compares the performance of two pinhole diameters. One offers more contrast and perceived sharpness, while the other provides more detail and resolution. (MTF data courtesy of Kjell Carlsson)

In 2004, Kjell Carlsson of the Royal Institute of Technology in Stockholm, Sweden conducted an evaluation of a variety of pinhole sizes. Unique to his approach was the fact that he stayed clear of subjectively comparing photographs. Instead, he computed MTF data for a number of different pinhole diameters and compared their MTF graphs. Fig.11 shows an example comparing the two proposed pinhole apertures. The diameter of equation (1) is derived from the Airy disc, and the diameter of equation (2) is based on the Rayleigh criterion. The comparison illustrates the performance difference of the two formulas, but it also reveals why an agreement for the optimal pinhole diameter was so difficult to achieve. Equation (1) offers more contrast and perceived sharpness, while equation (2) provides more detail and resolution.

fig.12a-b The test images in a) were taken with a small pinhole, based on the Airy disc, and the images in b) with a large pinhole, based on the Rayleigh criterion. The small pinhole in a) offers more contrast, while the large pinhole in b) provides more resolution. Most observers, however, perceive the high-contrast images on the left as being sharper of the two sets.

A set of test images in fig.12 verifies the theoretical evaluation. A small-format digital SLR (see fig.8) was equipped with a small pinhole, based on the Airy disc (0.25 mm), to create the images in fig.12a, and a large pinhole, based on the Rayleigh criterion (0.30 mm), to create the images in fig.12b. The images in fig.12a have more contrast and appear to be overall sharper than the images in fig.12b, as seen in the license plates, while the images in fig.12b have more resolution, as the bar charts reveal. Confusingly, this leaves us with two options for an optimal pinhole diameter, one for contrast and one for resolution. It is necessary to decide which of the two we want to optimize, before we agree to just one optimal pinhole diameter.

The quest for the optimal pinhole diameter is generally fueled by the desire to create the sharpest pinhole image possible. Contrast and resolution are both aspects of sharpness, but as demonstrated in fig.12, human perception typically prefers high-contrast images to high-resolution images. Consequently, unless resolution is more important than perceived sharpness, my proposal for the optimal pinhole diameter (d) is based on George Airy’s diffraction-limited disc:

or in the more conventional format:

where ‘λ’ is the wavelength of light, and ‘f’ is the focal length of the pinhole. A common value for the wavelength of light is 555 nm (0.000555 mm), which is the eye’s sensitivity peak and an appropriate value for standard pictorial photography. For infrared photography, use the film’s spectral sensitivity instead.

The graph in fig.13 shows how the optimal pinhole diameter increases with focal length, and the table in fig.14 provides useful data for some popular focal lengths to help with the design, exposure and composition of pinhole images.

Pinhole Aperture, Exposure and Focus

As we saw in fig.5a, regular sewing needles are convenient tools to create quality pinholes. Since the beginning of the 19th century, needle sizes are denoted by numbers, and the convention is that the thickness of a needle increases as its number decreases. In other words, the higher the needle size number, the thinner the needle. Fig.14 identifies the most appropriate needle size to create a popular pinhole diameter.

Fig.14 also shows the approximate pinhole aperture in f/stops with 1/3-stop accuracy. Use this aperture for all exposure calculations or measurements, and don’t forget to consider film reciprocity, as exposure times are likely long enough for reciprocity to have a significant effect. Most general-purpose lightmeters do not have aperture settings beyond f/64. This makes their application somewhat cumbersome for pinhole photography, where apertures of f/256 and smaller are the norm. However, fig.14 provides exposure compensation for all f/stops in relation to f/64. Set your lightmeter to f/64 to determine the exposure, and extend the exposure time according to the indicated f/64 compensation for your pinhole aperture. You will find a special pinhole dial in the appendix under ‘Tables and Templates’ to simplify this task.

fig.13     The optimal pinhole diameter for perceived sharpness is based on the equation for the Airy disc.

Most pinhole cameras do not provide any type of focus adjustment, and therefore, a pinhole camera is always focused at infinity. This means that the depth of field extends from the hyperfocal distance to infinity, and the hyperfocal distance is the front focus limit. A look at the hyperfocal distance in fig.14 demystifies why pinhole cameras are considered to have almost endless depth of field. At f/256 pinhole focus amazingly extends from 270 mm to infinity.

fig.14     This table provides useful data for some popular focal lengths to help with the design, exposure and composition of pinhole images.
a)   optimal pinhole diameter
b)   needle number to make pinhole
c)   working aperture in 1/3 stops
d)   exposure compensation relative to f/64 exposure measurement
e)   maximum pinhole resolution
f)   hyperfocal distance
g)   pinhole extension required to focus at hyperfocal distance

Depth of field can be extended even further if the pinhole camera provides some kind of a focus adjustment, as it would in a view camera conversion. Maximum depth of field is obtained when the pinhole is focused at the hyperfocal distance, in which case, depth of field starts at half the hyperfocal distance and extends to infinity. Of course, visual focusing is impossible with small pinhole apertures and the dim images they create. That is why the last column in fig.14 provides a dimension for the pinhole extension. Extend the pinhole-to-film distance by this amount in order to focus the image at the hyperfocal distance. As with all close-up photography, moving the pinhole closer to the subject moves it away from the film, which reduces film illumination. This must be compensated by an increase in exposure time, and in case of the optimal pinhole diameter, by an exposure increase of 1 1/6-stop for hyperfocal focusing.

fig.15     In the chapter ‘How to Build and Use a Zone Dial’, a useful Zone System dial is presented for general exposures. Pinhole photographers will be happy to know that they can find a special pinhole version in the appendix under ‘Tables and Templates’.

Pinhole Alternatives

There is hardly another field in photography more inviting to experimentation than pinhole photography, and modifying the pinhole aperture is a creative method to produce endless possibilities for image alternatives. If the aim is image clarity, a plain circular hole of optimal diameter is hard to beat, but if you like to explore unconventional substitutes, try apertures of all shapes, including horizontal, vertical and wavy slots. More technical aperture alternatives for pinholes are diffraction zone plates and photon sieves.

Lenses produce images through refraction; pinholes produce images through diffraction. With zone plates and photon sieves (fig.16), photographers take full advantage of diffraction by creating apertures that simulate the Airy diffraction pattern. Both have larger apertures and require less exposure than plain pinholes but produce fuzzier images with less depth of field.

fig.16     Diffraction zone plates and photon sieves are alternatives to a plain pinhole. They have larger apertures and require less exposure but produce fuzzier images with less depth of field.

A zone plate (fig.16b) consists of a center hole, which has the same diameter as the optimal pinhole, and an arbitrary number of concentric rings or zones, alternating between opaque and transparent. The outer diameter for each zone (dn) is given by:

where ‘λ’ is the wavelength of light, ‘f’ is the focal length of the pinhole, and ‘n’ is the sequential number of the zone. It is important to note that each zone, whether opaque or transparent, has the same surface area as the center pinhole. This means that a zone plate with seven additional transparent zones has eight times the light-gathering power of the pinhole alone, which is equivalent to an aperture improvement of +3 stops.

Another pinhole alternative is a multi-pinhole pattern, also called mega-pinhole or photon sieve. Instead of using the entire ring of a diffraction zone, as in the zone plate, an arbitrary number of small pinholes are distributed along the theoretical zones of the photon sieve, forming a hole pattern for each diffraction zone. While the diffraction zones become thinner and thinner as they ripple away from the center pinhole, the pattern holes become smaller and smaller towards the outside of the photon sieve. The design in fig.16c distributes just enough holes in each zone to equal half the surface area of the center pinhole for each hole pattern. This means that a photon sieve with six additional hole patterns has four times the light-gathering power of a single pinhole alone. This is equivalent to an aperture improvement of +2 stops.

Of course, it’s impossible to cut or drill zone plates and photon sieves like pinholes. The best way to make them is to create an enlarged, tone-reversed drawing of the design and photograph it onto high-contrast B&W film thus reducing it to the right size. Two design patterns are available in the appendix under ‘Tables and Templates’. The trade-off for increased light-gathering power with zone plates and photon sieves is a reduced depth of field and a loss of image quality, which is a result of larger apertures and less than perfectly transparent materials. Nevertheless, for many photographers, the unique image characteristics of these special apertures more than make up for all their disadvantages. The same is true for pinhole images in general. They are well worth a try.

 

Basics of Digital Capture

The essential elements of digital imaging, quality and archiving

This book predominantly covers the details of traditional darkroom work, because we believe that analog photography provides the most valuable final product possible: a silver-gelatin print, properly processed to archival standards. With the recent advent of digital imaging, however, even the most sophisticated image manipulation techniques are readily available to anyone with access to a powerful computer and specialized image software. Digital image manipulation is often easier and more powerful than its darkroom counterpart and typically delivers seamless results in less time. By combining analog photography and digital imaging, these sophisticated options also become available to the analog darkroom enthusiast. This chapter is an introduction to digital imaging in order to take advantage of these cross-over technologies, some of which are presented throughout the rest of the book. Digital imaging is a vast subject, which has already filled many books on its own. Consequently, we will not get into the intricacies of digital image manipulation, but we will introduce essential digital elements and discuss choices that have a direct bearing on protecting digitally stored image data and achieving the best image quality possible.

fig.1       The color original of this image was taken with a digital SLR and converted to monochrome through imaging software. To ensure these delicate white flowers show plenty of detail, the image was taken in diffuse sunlight with a small degree of underexposure and using a tripod. An aperture of f/11 was used to ensure all the petals were in focus, though at this small aperture setting, the image quality was already starting to be limited by diffraction. Digital equivalents of traditional darkroom manipulations were used to suppress edge detail and lift the tonal values.

fig.2       The Canon EOS 5D is one of the world’s first full-frame digital SLRs, featuring 12.8 million effective pixels at a pixel size of 8 µm (microns).

(image copyright Canon, Inc.)

fig.3       The full-frame sensor of a Nikon D3x provides 24.5 million pixels at a size of 6 microns each.

(image copyright Nikon, Inc.)

Digital Camera Sensors

Because film is a relatively cheap consumable, we tend to forget the amazing technology behind it. Simply put, film is a plastic strip, coated with a thin layer of gelatin and loaded with light-sensitive silver salts. Understanding the boundaries of this remarkable commodity is the key to its full exploitation. Similarly, in the fast moving world of digital imaging, it is essential to understand the basic function, a few essentials and the physical limitations involved with digital sensor design to make use of their full potential. For both film and digital systems, there is no magic formula. In spite of continuous technological advancement in digital imaging, there are still considerable trade-offs between cost and image quality, not to mention the ultimate limits placed on digital capture by the laws of physics.

fig.4       To be useful for digital imaging, image detail must be recorded in samples small enough to be unidentifiable as a matrix of pixels when the final print is observed from a normal viewing distance. From left to right: this image was recorded to be shown at 12, 60 and 300 ppi (pixels per inch). The matrix of pixels is very obvious at 12 ppi, and from a minimum viewing distance, it is still clearly detectable at 60 ppi. At 300 ppi (equivalent to 6 lp/mm), however, the digital origin of the image is nearly concealed.

Sensor Elements, Pixels and Resolution

A simple photoelectric sensor transforms light energy into an electrical signal. To make this process useful for digital imaging, the analog signal is converted into a numeric value using an analog-to-digital converter, often called A/D converter or simply ADC.

To record an entire image digitally, one needs a closely packed array of sensor elements, whose signals are converted by the ADC into an orderly sequence of numbers. During the camera exposure, each sensor element collects and stores the energy from the photons they receive. The camera electronics then measure the captured energy level for each sensor element and convert it, with the help of the ADC, into a matrix of distinct intensity levels (fig.4). In this way, digital cameras scan or sample the image in fine increments and record them as image detail. Generally speaking, the finer the sample increments are, the more realistic the final digital image appears to the viewer. Image resolution must be fine enough to be unidentifiable as a matrix of pixels when the final print is observed from a normal viewing distance.

Unlike film, which is a homogenous photosensitive surface (see fig.5), digital camera sensors do not actually have sensor elements covering the entire surface area of the array. In some cases, they cover just half the image sensor surface in order to accommodate the supporting electronics in-between them. But, discarding light energy is wasteful and forces the electronics to work with a weaker signal. Digital cameras minimize this problem by placing a microlens above each sensor element to enhance their light-gathering ability. This improves the image sensor efficiency and signal strength of each sensor element.

fig.5       Unlike film (left), which is a homogenous photosensitive surface, digital camera sensors do not have sensor elements covering the entire surface area of the array, in order to accommodate the electronics in-between them.

The sensor pitch is the physical distance between two sensor elements and is equal to the effective pixel size. Typically, current digital SLRs have an effective pixel size of about 5-8 µm (microns). Compact digital cameras and mobile phones often offer the same mega-pixel count but on a much smaller sensor array. As the pixel size is reduced, either as a result of the overall sensor shrinking, or from packing more pixels into the same sensor real-estate, the light gathering ability of each sensor element is also reduced. As a consequence, the sensor resolution is improved by the number of pixels, but the signal level of each sensor element is lowered. The ongoing challenge is to design image sensors with higher packing densities without compromising the optical efficiency. The current state of technology suggests that the optimum pixel size is around 7-8 microns, leading to the conclusion that better resolution and overall performance can only be achieved by increasing the sensor size, and not by reducing the pixel size.

fig.6       An additive color system starts with no light (black) and adds the three primary colors Red, Green and Blue (RGB) in varying amounts to produce any color of the visible spectrum. Combining all primary colors in equal intensities produces white.

Nevertheless, the trend in digital sensor design is to increase the pixel count. Some increases are more meaningful than others. As long as the size of the image sensors remains unchanged, every doubling of the amount of pixels increases the sensor resolution by more than 40%. A change from 10 to 12 megapixels increases resolution by less than 10%.

To satisfy the criteria of standard image resolution, one needs at least 370-image ppi (pixels per inch) for a 5x7-inch print, because a print this small is typically observed from the closest possible viewing distance. As print sizes and viewing distances increase, resolution requirements are reduced proportionally. A 16x20-inch print needs as little as 140 ppi to look convincingly realistic, and a billboard across the road may need no more than 12 ppi to conceal its pixelated origin.

Color Perception

Image sensors are essentially panchromatic, although they exhibit a varying sensitivity to different wavelengths of light, just as film does. Color science differentiates between additive and subtractive color systems. An additive color system starts with no light (black) and adds the three primary colors Red, Green and Blue (RGB) in varying amounts to produce any color possible in the visible spectrum (fig.6). Combining all primary colors in equal intensities produces white. This creates the opportunity to measure image color by combining the results of three sensor elements, which have been made color selective through individual color filters.

fig.7       The Bayer array takes into account that human vision is particularly sensitive to green light, and features twice as many green filters, as red and blue filters. Each pixel captures only one primary color, but ‘true’ color is calculated from neighboring pixels.

It may appear that distributing image color measurement to three different sensor elements comes at the expense of reduced image resolution. In nearly all digital cameras, this problem is solved through an ingenious pattern of color filters (fig.7) called a ‘Bayer array’. In this array, a group of four pixels, each containing one red, two green and one blue filter, is assigned to collect one full piece of color information. This pattern takes into account that human vision is particularly sensitive to green light, and features twice as many green filters as red and blue filters. Each filter is located directly on top of a sensor element, so that each pixel captures only one color channel. The missing channels for each pixel are calculated from neighboring pixels through a process called ‘demosaicing’. The color image is recorded in the form of an RGB file, which is made up of three color channels and contains ‘true’ color information for each pixel on the image sensor.

Moiré

When two regular patterns of closely spaced lines are superimposed, they create another pattern of irregular wavy lines, called moiré (fig.8). The image sensor’s closely spaced array of pixels is organized in a regular pattern. If the subject to be photographed also contains a regular, closely spaced pattern, then disturbing moiré lines may be observed in the picture. Common subject details, such as the shingles on a roof, a distant fence or some fabrics, for example window curtains, are prone to this effect. To minimize the problem, many cameras are equipped with a mildly diffusing moiré filter in front of the sensor.

fig.8       When two regular patterns of closely spaced lines are superimposed (top), they show a pattern of irregular wavy lines, called moiré. The image sensor’s pixel pattern, in combination with certain subjects (bottom), may create moiré lines in digital photographs.

Noise

Ultimately, the quality of any device is limited by the small difference between the signal transmitted and the signal received. The words heard over the phone are never quite as clear as the words spoken at the other end. Analog and digital cameras have a similar limitation. With an analog camera, the film grain limits the level of fine subject detail the camera can capture, and the digital camera equivalent of film grain is called image noise.

fig.9       Two images of a uniform surface were taken with a digital compact camera and a professional digital SLR at low light with a high ISO setting. These 300 ppi examples clearly show the advantage of larger image pixels.

Each sensor element transforms the light energy received into an electrical signal, which is converted into a numeric value by the analog-to-digital converter. If the sensor element is struck by a bright highlight, the signal is strong, and if the light was transmitted by a dim shadow detail, the signal is weak. Unfortunately, sensor technology is not perfect, and while every sensor element transforms the light energy received into a signal, it also adds some random noise or sporadic peaks. For the most part, identical light levels are transformed into slightly different signal strengths by different sensor elements, which are then converted into varying numeric values by the ADC. The result is a more or less constant image noise, which appears as random speckles on an otherwise uniform surface (fig.9).

Image noise appears predominantly in areas of low exposure and shows up most disturbingly in smooth tones. Noise is amplified with higher ISO settings and longer exposures, but is less problematic with larger sensors, because large sensors have large sensor elements that collect more light and create stronger signals than small elements. This means that the sensor noise is only a small fraction of the sensor signal. The signal-to-noise ratio (SNR) is a useful and universal method to compare the relative amounts of signal and noise in any system. High signal-to-noise ratios will have very little apparent image degradation whereas the opposite is true for low ratios. High-quality sensors aim to make the noise level insignificant with respect to the signal, so that its influence is minimal.

Speed

A film’s ISO speed describes its sensitivity to light. Digital cameras can uniquely capture images at many different ISO speeds. This is accomplished by amplifying the sensor signal prior to the conversion into a digital number. Amplification does not improve the signal-to-noise ratio, since it amplifies the combined sensor signal and noise equally. As a consequence, camera exposures at high ISO speeds, which capture low light energy levels, will produce significant image noise. For the best image quality, one should select a low ISO value, use the optimum aperture and support the camera with a tripod.

Depth of Field and Resolution Limits

Broadly speaking, for a given aperture, the depth of field for a lens is inversely proportional to the focal length. Therefore, optics with a short focal length offer more depth of field than longer lenses. Small camera formats require shorter focal lengths in order to provide the same angle of view than larger formats and have a larger depth of field at similar aperture. This is one reason why small digital compact cameras have such an enormous depth of field. In other words, digital cameras do not offer more depth of field than film cameras, but a small camera format offers more depth of field than a large camera format, because it typically uses lenses with shorter focal lengths.

With film cameras, image resolution is limited by lens aberrations and diffraction alone (see ‘Sharpness and Depth of Field’, fig.11 for details). Regular film is not a limiting factor, because the resolution potential of its fine grain is above the combined limits of aberrations and diffraction. However, with digital cameras, the resolution of the image sensor cannot be ignored. At working apertures, sensor resolution is typically the only limiting factor of digital image resolution.

As a wide-open digital camera lens is stopped down, image resolution increases at first, because lens aberrations are reduced. Image resolution peaks at an ‘optimal’ aperture limited by sensor resolution. Stopping the lens down further decreases image resolution again, due to the ever increasing influence of diffraction, but it requires very small apertures (f/22 or smaller) before diffraction becomes the only limiting factor of image resolution.

For any depth of field calculation this means, if the sensor resolution (Rdigital) is coarser than the circle of confusion required to support the viewing conditions, the optical system is limited by the sensor, and the smallest circle of confusion (cmin) is given by:

As an example, if the image sensor has a pixel size of 8 microns, and at least 2.1 pixels are needed to reliably record a line pair, then sensor resolution is 60 lp/mm (1/(0.008x2.1)), and there is no need to take a smaller circle of confusion than 0.017 mm (1/60) into account, because the sensor resolution does not support it.

Tonal Control

Photographers working in the digital domain enjoy a remarkable advantage to the envy of every darkroom worker. This is the ability to manipulate image tonality almost endlessly. At the simplest level of digital image manipulation, the overall contrast and tonal distribution can be averaged or adjusted to preset standards. At its most sophisticated level, digital image manipulation permits overall or local image tonality to be precisely controlled using a variety of specialized creative tools. Three tools, however, accomplish the majority of tonality control: ‘histogram’, ‘levels’ and ‘curves’.

fig.10     The histogram, typically provided with digital cameras and imaging software, is a common tool to quickly analyze the distribution of brightness values and image tonality.

Histogram

A histogram is an efficient graphical method to illustrate the distribution of large data sets. Typically provided with digital cameras and imaging software, the histogram is a common tool to quickly analyze the distribution of brightness values and, consequently, image tonality. Fig.10 shows an example of a histogram on a digital camera and as a feature of imaging software. Both follow the same principle.

The horizontal axis represents all image tones from black (left) to white (rihgt), and the vertical axis represents the relative amount of pixels using each tonal value. At a glance, this visual aid indicates whether an image uses the available tonal range, is generally under- or overexposed, and whether the exposure is clipped, losing essential shadow or highlight information. Ideally, the response should tail off, just before reaching the extreme ends of the scale. The histogram is often used in conjunction with tonal controls such as ‘levels’ and ‘curves’.

fig.11     The most common tools for tonal control are ‘levels’ (a) and ‘curves’ (b).

Levels and Curves

Nearly every digital image requires some change to exposure and contrast to improve tonality. The most common tonal controls are ‘levels’ and ‘curves’, which are present in all sophisticated imaging software. Of the two, ‘curves’ is the most powerful, whereas the ‘levels’ adjustment has a simpler interface and a reduced flexibility of tonal control. Either way, the effective contrast and brightness of an image is changed so that key highlight and shadow areas have the correct tonal values. This is then confirmed by placing the eyedropper tool into these key areas and reading the RGB or grayscale information at that point.

fig.11a shows a typical ‘levels’ dialog box. It includes a histogram of the image, immediately above three sliders. The two outer sliders effectively control the shadow and highlight endpoints of the image, much like basic exposure and contrast controls in darkroom printing. Moving these sliders towards the center increases image contrast, but if they are moved into the histogram distribution, some image tones are clipped into featureless black or white. The third slider in the middle controls the tonal distribution smoothly between the endpoints, effectively lightening or darkening the midtones. The darkroom equivalent to this is more involved, because it requires switching to a different film or paper, or modified processing. If an image looks good on-screen but suffers from empty highlights and blocked shadows when printed, then moving the bottom two sliders towards the center lowers the contrast and redistributes the image tones evenly between the printable tonal extremes.

fig.12     This sequence shows how increasing bit-depth ultimately provides photo-realistic images. From top to bottom, 1, 2, 3, 4 and 6 bits per pixel allow for 2, 4, 8, 16 and 64 levels of gray.

Fig.11b shows an example of the more sophisticated ‘curves’ adjustment tool. In essence, it allows the user to map any tone to any other tone using a transfer curve. In doing so, one can change exposure and contrast, or create nonlinear tonal distributions and control highlight or shadow separation at the same time. It can be used to mimic camera filters, complex darkroom techniques and much more. The example shown here includes the histogram in the background for reference and uses a gentle S-curve tonal adjustment to increase midtone contrast. The curve can be adjusted by numerical input or arbitrarily reshaping the curve with the mouse.

Both the ‘levels’ and ‘curves’ adjustments can be applied to the entire image or only to a selection. Both tools can do far more than can be explained here in a few paragraphs, and beyond this introduction comes the stony road of practice and experience.

Bit Depth

The bit depth of an image refers to the number of binary digits that describe the brightness or color content of an image pixel. A single binary digit is either on or off, and therefore, it can represent only the numbers ‘0’ or ‘1’. A black and white image without real grays can be described by a sequence of 1-bit digits, but to record intermediary levels, more tonal resolution and more binary digits are necessary.

Fig.12 compares a sequence of images, rendered at several low bit-depths. Beyond that, an 8-bit (1 byte) grayscale image has the potential to show 256 levels of gray, and a 24-bit RGB color image, with 1 byte or 8 bits for each color channel, can show over 16 million different colors.

The camera hardware determines the maximum bit depth, which typically ranges from 8-16 bits per channel. With 16 bits per channel, more than 65,000 levels of gray and over 281 trillion different colors can be stored. This may seem a little extreme, but they have been made available for good reason.

An experienced observer with good eyesight can detect the differences between roughly 200 evenly distributed gray levels and 10 million colors. Therefore, one might think that 8-bit image data per channel is more than sufficient for quality work, but this is not the case in practice, because we like to end up with evenly distributed image tonality after all image manipulations are completed. As we will see, this requires an abundance of image data to start with.

Posterization

In order for a photograph to look realistic, it must have an abundance of image tones with smooth tonal gradation between them. This requires an image file with sufficient bit depth. If the bit depth is too low, smooth tonal gradation is impossible, and what was meant to be a continuous-tone image is reduced to a limited number of gray levels or colors. If overdone, the loss of image tones becomes obvious and the image starts looking like a mass-produced pop-art poster and not like a realistic photograph. At that point, the image is ‘posterized’, and the process of reducing the bit depth to that extreme is called posterization.

The most common cause of posterization is extreme image manipulation through software tools such as ‘levels’ and ‘curves’. Posterization is more obvious in areas of smooth tonal transition, such as in skies, studio backgrounds, polished surfaces and smooth skin tones. These areas require delicate tones to describe them, and any decrease in bit depth can quickly have a visual impact. The best way to avoid posterization is to manipulate only 16-bit images or keep 8-bit manipulation to an absolute minimum.

A potential danger of posterization is easily detected by reviewing the image file’s histogram. Fig.13a shows the histogram of an 8-bit image file, which is obviously missing most midtone and all highlight values. Fig.13b shows the histogram of the same file after the tonality was spread out, in an attempt to obtain a full tonal scale image, and several other corrections were applied to optimize image appearance. Any gap in the histogram indicates pixel values without occurrence and, consequently, missing image tones. Small gaps are not necessarily causing posterization, but larger gaps are clear warning signs of potential posterization. Fig.13b indicates that the 8-bit image file did not have enough tonal information to support such extreme manipulation. The result is a posterized image, which is missing too many tonal values.

Fig.13c shows the histogram of a file that had been identically manipulated, but this time, the original image file contained 16 bits per pixel. The resulting image is not missing any pixel values and features a smooth tonal distribution from black to white. To illustrate the effect of posterization in actual prints, fig.14 shows two examples of image manipulation applied to an 8- and 16-bit image.

Posterization may also occur after converting an image from one color space to another. For monochrome work, the effect is minimized by recording exposures in the camera’s raw file format and converting them to 16-bit grayscale images before any manipulation attempt is made. If one must work with an 8-bit image, start by converting it to a 16-bit grayscale image and apply a minimal amount of Gaussian blur. This minimizes the possibility of posterization in subsequent editing. In any event, the histogram will always highlight any gaps in tonality.

fig.13     These histograms illustrate the effect of posterization. An 8-bit image file (a) was subjected to a number of rigorous tonal manipulations, which resulted in many unsightly discontinuities of tonal distribution (b), but in (c), where the origin was a 16-bit image file, this did not happen.

fig.14     An identical sequence of tonal manipulations were applied to these images. The 8-bit image (a) shows clear signs of posterization. The 16-bit image (b) shows no signs of gradation and features smooth and realistic image tones.

fig.15     High Dynamic Range imaging, or HDR, relies on blending two or more different exposures of the same scene. The exposures typically range from several stops of underexposure to several stops of overexposure. The dynamic range of an image can be extended beyond photorealism this way, reaching into surrealism.

Dynamic Range

The average photographic scene has a subject brightness range (SBR) of about 7 stops. In extreme lighting conditions, this range can be as low as 5 or as high as 10 stops. The dynamic range of an optical recording device is the maximum brightness range within which it is capable of obtaining meaningful data.

The human eye has an amazing dynamic range. The retina provides a static sensitivity range of about 6 stops. With the support of a light-regulating iris and quick selective viewing, the sensitivity range is extended to about 10 stops. Adding the ability to chemically adapt to a wide range of brightness levels, our eyes have a dynamic range of almost 30 stops.

With film and camera, we do not have the flexibility of selective viewing, nor do we have the time for brightness adaptation during an exposure, but if processed accordingly, film has an exposure latitude of 15 stops or more. Digital cameras, on the other hand, are particularly challenged by large subject brightness ranges. The dynamic range of today’s digital SLRs cannot compete with monochrome film and is typically limited to 7-9 stops. There are two solutions to improve matters considerably.

The first method is deceptively simple. A deliberate underexposure is made at a low ISO setting, so that the highlights are fully rendered and far from being clipped. This exposure is recorded at the highest bit depth possible and imported into the imaging software as a 16-bit file. Then, software image adjustments are made to lift the shadow detail and roll off the highlights, which effectively extends the dynamic range into the shadow region and lightens the midtones. Several extra stops of dynamic range can be gained this way. It is a technique used by wedding photographers to avoid overexposing the bride’s dress while still capturing the weave in the groom’s suit.

The second method relies on blending two or more different exposures of the same scene (fig.15). The exposures typically range from several stops of underexposure to several stops of overexposure. The dynamic range of an image can be extended beyond photorealism this way, reaching into surrealism. Sophisticated imaging software either supports combining the exposures manually, or it provides special features to automatically merge the exposures to one. This is called ‘High Dynamic Range’ or HDR. Creating a new image from two exposures, just a few stops apart, usually results in a realistic representation. More surrealistic images are made from several exposures covering an extreme subject brightness range. Every print has a finite contrast range, and simply squeezing in an unrealistic subject brightness range gives unrealistic looking results. This may serve to extend the boundaries of photographic creativity, but selective manipulation is the better choice if more convincing images are required.

Preparing for Digital Output

The last steps of digital image manipulation are sizing, scaling and sharpening of the image to optimize it for a specific output device. Every image file has a set number of pixels that control the image resolution on screen and in the final output. If this resolution is changed, the image expands or shrinks in physical size, since the total number of pixels remains constant. The pixel resolution, required for on-screen display or printers, is usually quite different. As a consequence, the image file may have either an excessive or an insufficient pixel count for its final purpose. Typical computer monitors feature resolutions of 65-130 ppi, whereas inkjet printers and half-tone imagesetters may require anything from 240-450 ppi. To support specific output requirements, the image file must be re-sampled to the correct output dimensions and the appropriate pixel resolution.

Re-sampling an image may create additional, or eliminate existing, pixels through a process called interpolation. This process requires that new pixel values be calculated from neighboring pixels in the original file through a number of alternative algorithms. Reducing the pixel count discards information and reduces image resolution. Conversely, increasing the pixel count does not increase resolution or add detail. It is important to make sure that there is sufficient resolution to support the intended print size before committing the data file to digital output. See the text box, earlier in this chapter, for some popular print sizes and their recommended image resolutions.

Sharpening

Due to all the mathematical acrobatics of generating new pixel values from neighboring pixels, and the use of optical anti-aliasing or moiré filters in front of the sensor, most digital images require some degree of sharpening. This is applied either within the camera software, right after image capture, or more controllably, at the last stages of image manipulation. The best practice is to sharpen the image just prior to reproduction. For that reason, professionals will keep a manipulated, but not sharpened, version of a premium image in addition to several reproduction copies. It is also a good practice to sharpen the image only where required. For instance, sharpening clouds and other areas of smooth tone has no pictorial benefit and may only accentuate image noise. This is another incentive to work exclusively with camera raw files and not to rely on in-camera sharpening for quality work.

Behind the scenes, the sharpening process involves re-calculating each pixel value again, based upon its relative brightness to neighboring pixels, always looking for opportunities to improve acutance. There are different sharpening tools available, all optimized for specific image styles, and each with its unique control settings. The most common and universal software tool is the so-called ‘unsharp mask’, which achieves mathematically the same optical effect as the darkroom process of the same name. Most applications provide a preview of the outcome (fig.16d), and one is well advised to evaluate the results, close to the final print scale, with the preview zoom level set to 100%, 50%, 25% and so on. Other zoom levels may create strange on-screen effects and disguise the sharpening effect. There are no ideal settings for the unsharp mask, since the optimum level changes with image resolution, size, noise and content. However, software sharpening is easily overdone, and less is often more.

fig.16a-c Soft images (a) are carefully sharpened (b) to restore their original brilliance. However, software sharpening is easily taken too far (c).

The examples in fig.16a-c compare different levels of sharpening. A slightly soft image (a) was sharpened with an unsharp mask, using the starting-point settings in fig.16d, which successfully improved image sharpness (b) and restored the original subject brilliance. A much stronger setting exaggerated image contrast (c), and delivered an unsightly print, similar to what we get from the office copy machine.

fig.16d    The ‘Unsharp Mask’ dialog box in Photoshop offers three main controls, which affect the level of sharpening, the spread of the mask and a threshold to avoid accentuating image noise. The settings shown here are a good starting point and also avoid over-sharpened and unsightly images.

Imaging File Formats

The image file is the digital equivalent of a traditional negative, and as such, it is considered to be the image original. As with film negatives, digital image files deserve the utmost care, otherwise, the original image is lost forever. The first consideration is usually the choice of format in which the image file should be stored. This initial choice of file format defines the limits of digital image compatibility and quality.

We differentiate between uncompressed and compressed file formats. File compression is used to reduce the file size of the digital image, and is either lossless or ‘lossy’, in which case, image quality is likely to be compromised to some degree. Lossless compression schemes simply eliminate data redundancies, and it is very effective with images that contain large homogeneous areas or repetitive patterns. With other images, the data reduction is insignificant, and the file size may even inflate. Lossy compression algorithms eliminate image information considered to be of little interest to the viewer, and file sizes vary according to compression factor. However, once image information is lost, it cannot be brought back.

Several file formats dominate the consumer and professional markets. It is wise to preserve not only the manipulated version of an image, but the original camera image as well, so that one can take advantage of improved editing with the latest software.

fig.17     Image files are stored in uncom-pressed or compressed formats. Some algorithms eliminate image information considered to be of minor significance, according to preset compression levels. However, once image information is lost, it cannot be brought back.

JPEG (.jpg)

This image file format was created by the Joint Photographers Experts Group and is the most compatible, highly file-size efficient and well-established lossy compression scheme. JPEG files are often found as the default format in consumer digital cameras and as the ‘snapshot’ alternative in professional SLRs. JPEG files support RGB and CMYK color spaces, but are limited to 8 bits per channel. They are a compromise in quality, often leading to unsightly artifacts at high compression rates. They are not a good choice for extensive image manipulation or high-quality work.

TIFF (.tif)

The Tagged Image File Format, or TIFF, can be un-compressed or compressed but is always lossless and records 8, 16 or 32-bit grayscale, RGB or CMYK images. The format has developed into a stable standard, is widely compatible with desktop publishing software and popular with professional printers. It also records additional image layers, masks and paths. Some digital cameras can produce TIFF files directly.

Photoshop (.psd)

Photoshop files are included here, simply due to the dominance of Adobe Photoshop in the marketplace. Photoshop’s file format is well compressed and lossless, while allowing for a maximum degree of editing flexibility and compatibility with other formats, as long as you own Adobe Photoshop. The file format supports an assortment of color spaces, and just like TIFF, preserves image layers, masks and paths, at the expense of file size. This overcomes the limitations of Photoshop’s destructive editing nature. Photoshop is frequently updated, and its feature set is always improved and extended. However, version control is required to maintain backwards file compatibility.

Camera Raw (.nef, .cr2, .orf, …)

File formats in this class record image data directly from the camera sensor with a minimum of in-camera processing. In-camera processing is a compromise to maximize speed and lower power consumption, and is best kept to a minimum. The format is used to store high-bit-depth RGB images, and allows the user to tune exposure, sharpness, contrast and color balance, to name just a few image characteristics. Camera raw files have the best potential to produce high-quality images. Unfortunately, each manufacturer has a proprietary camera raw format, which is upgraded with new camera releases. This often demands computer software updates and is exploited by some companies to force upgrade purchases. Early attempts to standardize raw formats have failed, and it is not certain that older formats will remain supported in new operating systems and applications. In view of this, our recommendation is to always start with a camera raw file, but archive original image files as either ‘Digital Negatives’ or as high-quality, 16-bit TIFF files, which are, unfortunately, very demanding of space.

Digital Negatives (.dng)

This image file standard is an open format, created by Adobe, in an attempt to overcome the transitory nature of camera raw files. It is meant to be an archival format, containing the essential image data of proprietary camera raw files. It is lossless but highly compressed, and by creating this industry standard, it is hoped that obsolescence will be reduced. Major consumer brands will be the last to desert their proprietary formats, especially while improvements and software enhancements are still numerous. In view of this, it is comforting to have an open image format, which promises to support longer-term archiving.

Archival Storage

In order to store digital image files safely for a long time, it is worth considering a few obstacles. In addition to its limited life expectancy, restricted through physical and chemical deterioration, digital media also faces the problem of obsolescence. Media obsolescence is most frustrating. Nothing seems to be immune from the march of technology. One may have stored images on reliable recording media under the best environmental conditions, only to find out that hardware interfaces, software applications, operating systems or recording media have changed yet again, and that there is no easy way to get to the image data anymore. The new software version cannot read the old files, the old hardware interface is incompatible with the new computer or the new hardware does not accept the old storage media. One’s only defense against both deterioration and obsolescence is to transfer digital image files occasionally from old to new storage media with or without an upgrade in technology.

However, the life expectancy of electronic media and its time to obsolescence vary greatly. Removable media in the form of optical disks, as in CDs and DVDs for example, are prone to physical and environmental damage. As with print storage, temperature, humidity, exposure to light, handling and airborne oxidants all contribute to early failure. In addition, the material choices, laser power, write-speed and the manufacturing variations between disks affect the longevity of the media as well.

There are three main categories of optical disks: read-only, recordable and rewritable. They all have a polycarbonate plastic substrate but differ in data-layer technology. Read-only CD/DVD-ROM disks are by far the most reliable, because the data is molded into the disk as a spiral track of pits, similar to the grooves in audio records, and a laser reads the digital information from the pits. Unfortunately, their manufacture requires industrial-size machinery, which prohibits their use for individual image files. Recordable CD/DVD±R disks use a gold or silver layer, coated with an organic dye, to store the data. As the laser records the information, the dye becomes discolored, which encodes the information. The gold variety of disk is very durable, and their low-light life expectancy is assumed to be 20 years or longer. Rewritable CD/DVD±RW and RAM disks, on the other hand, use a metal-alloy film on aluminum, which is subject to premature oxidation and, therefore, not recommended for long-term storage.

It must be mentioned that the organic dyes used in recordable media are very sensitive to UV radiation and will deteriorate within days or weeks in strong sunlight. The Optical Storage Technology Association (OSTA) state that it is extremely difficult to estimate expected disk life, but suggest that the shelf life of unrecorded media may only be from 5-10 years. It also casts doubt on some manufacturers’ claims of optical disks lasting up to 50 years. There is an ISO standard for accelerated testing, but it only considers temperature and humidity variation. Poor technique, manufacturing variability, recorder settings, handling and storage method will significantly reduce data life expectancy. Several institutions have reported their disks failed within 2 years, for no obvious reason. Some, consequently, have switched to magnetic storage methods, such as hard disks, magneto optical disks or tape systems for long-term backups. Indeed, an external hard disk, used for backups, may be an ideal long-term solution, but many users prefer to use optical media for convenience and economy.

fig.18     In addition to its limited life expectancy, digital media also faces the problem of obsolescence.

Best practice mandates high-quality materials, optimum recording media and image data storage under ideal conditions. First, make sure that your hardware and software is up-to-date, which ensures that it matches the capabilities of the latest media. Second, choose gold CD/DVD±R disks, advertised as archival, handle them at the edges only, and record at slow data rates. Verify the disk after recording, and store it vertically in an inert, acid-free sleeve, in a cool, dry and dark place. Use a common file format and avoid the proprietary formats found with back-up programs to avoid future compatibility issues. (For example, the Windows 2000 operating system cannot read the Windows 98 backup file format and, worse, modern PCs cannot load Windows 98 or Windows 2000, because they do not have the necessary hardware drivers).

It is essential to name digital files descriptively and catalogue archives, because it is all too easy to lose an image in the metaphorical haystack, without being able to search, find and see its metadata. Several commercial software programs are available to logically store, effectively search and quickly retrieve digital image files. Finally, each time you update equipment or software, it is advisable to check the compatibility of your archives with the new hardware and software before disposing of the old equipment.

Against this backdrop of uncertainty, it is worth remembering that, as long as there are optical systems, monochrome film negatives have no real media obsolescence. They can be projected, copied, scanned and simply investigated with a loupe. Their chemical and environmental deterioration is well understood and relatively easy to control within known limits. Film and paper have a proven track record of over 150 years in real-world conditions, and have no need for questionable advertising claims, which are at best derived from accelerated testing. We like to think that the best of our creative efforts will last. While negatives are sitting patiently, unattended and just waiting to be discovered, digital image files might last for a long time, but their dormant bits and bytes are likely to be unreadable by future generations without constant checking and re-recording. Who knows, the best way to preserve digital images may be to convert them to analog files! A currently popular option is to print them with archival inks on archival paper and keep them in a cool, dry and dark place before they irretrievably fall into what some image conservationists refer to as the ‘digital gap’.

 

Digital Capture Alternatives

Comparing and choosing solutions for digital monochrome

The roots of this book are planted firmly in the traditional domain. Despite the allure and advances made by digital cameras and printers over the last decade, nothing approaches the beauty, permanence and depth of a toned, fiber-base print. Given that an image may not have been initially intended as a traditional monochrome print, or requires manipulations that are most efficiently performed digitally, this chapter compares the alternative methods necessary to bring an image, either directly or indirectly, into the digital domain for the purpose of editing and final output onto silver-based photographic paper. Clearly any recommendation will be challenged by evolving technology, and so, the assessment criteria and methods are explained, to be reevaluated in the user’s own time.

Imaging Paths

The diversity of available media, printing methods and imaging equipment make the many and varied routes from subject to final image worth contemplating. Disregarding web images for the moment, fig.1 shows an overview of the possible imaging paths from subject to print. Of interest, here, are the highlighted items, which bring an image into the digital domain for editing and still allow a full range of output options into analog printing. When deciding on the capture method, apart from the immediate issues of recording the subject satisfactorily, it is also necessary to consider the demands of downstream requirements, which are related to print size, printing method, archival requirements or accepted media.

Clearly, there are two main starting points for imaging a digital file: a) indirectly through film-based systems or b) directly from a digital camera. Our comparisons between analog and digital systems are made without reference to specific models. We have instead referred performance to quoted specifications, which allows the reader, up to a point, to infer the performance of future digital equipment. Clearly, any conclusion is wholly dependent upon the relative importance of an image’s quality parameters, and, without some conscious prioritization, can be the subject of endless debate. It is not uncommon for protagonists to infer superiority of one path over another, based on a single parameter and conveniently ignore others. Their relative importance also changes with consumer trends. Especially, quality and longevity are disregarded over the marketed appeal of new technology. The relative importance of these parameters also varies with the intended imaging purpose. For instance, a significant advantage of a digital camera is its ability to adapt to the ambient light color temperature, but this is, of course, of little value for monochrome work.

In these assessments, we assume that the underlying purpose is always to make a monochrome print on silver-based photographic paper with qualities suitable for a fine-art landscape photography. We cover aspects that are directly measurable, as well as our subjective evaluations. Each of our readers should consider their own priorities and shuffle the following parameters in order of importance and according to taste, image style and application. Beyond these considerations is the truism that any camera or photo is better than no camera or a missed shot.

fig.1       There are many ways to get from image capture to the final print. In this chapter, we compare several digital capture alternatives to film, in order to explore the limitations of producing a high-quality, silver-gelatin print from digital capture.

Quality Parameters

Dynamic Range

The chosen capture system should be able to record the required subject brightness range and ensure a full-range print can be made with sufficient highlight and shadow detail. Digital and film media have an inherent capability, both of which can be enhanced, to some extent, by software and darkroom controls, respectively. The dynamic range of an optical recording device is the maximum brightness range within which it is capable of obtaining meaningful data.

Resolution

Another significant consideration is the effective system resolution. In the chapter ‘Sharpness and Depth of Field’, we define the closest comfortable viewing distance for a print at about 250 mm and the standard viewing distance as approximately equal to a print’s diagonal dimension. Human vision can, in general, resolve about 7 lp/mm on a print at 250 mm. As the viewing distance increases, print resolution can be lowered without obvious detection, although humans can distinguish prints with higher resolution beyond their physiological limit. The imaging system should meet or exceed the performance threshold of standard human vision. The requirements for film or digital media vary with format. Fig.2 lists resolution requirements and sampling rates needed to effectively capture them at two MTF contrasts, 10 and 50%, which imply the limit of resolution and acceptable sharpness.

Tonality

Once any image, irrespective of the source, is in the digital domain, the subjective distribution of tones between highlights and shadows is under the direct control of the imaging software. Extreme tonal manipulations require images with low noise and a high bit-depth, as camera raw files and 16-bit film scans. For monochrome work derived from digital color originals, there is an interesting twist, because when the starting point is a color original, one can change the monochrome tonality by employing filtration, just as one does on-camera with monochrome film.

Sharpness, Grain and Noise

These attributes are intentionally grouped together, since a significant improvement in one often causes obvious deterioration in another. Images obtained by scanners or digital cameras are best captured with minimal sharpening settings and then sharpened to the required level in the imaging software. Sharpening algorithms amplify image grain and noise, and can also dramatically change image appearance. Conversely, image noise or grain can be reduced with digital blurring filters, at the expense of image sharpness and resolution. More advanced digital filters and plug-ins exist, and they are constantly evolving to intelligently minimize image degradation. However, one ought to fully understand the existing software filters before reaching for costly alternatives.

A note of caution: the initial visual appeal and ease with which a digital image may be sharpened often leads to its over-application. While image sharpness may hold sway over image resolution, its over-use destroys tonal subtlety and resolution. It is important to be aware of the balance and interaction of the sharpness controls with unwanted image side effects for each proposed system and make one’s own assessment of the optimum balance. Although most individuals can effectively compare side-by-side image sharpness, a proposal is made later on for an absolute measure.

Pause a moment to qualify the above mentioned quality parameters and consider your own print-making experiences. For instance, although fine images often require high levels of resolution, it may be possible to work with less, based upon image content and viewing conditions, especially if the image has simple shapes, is noise-free and sharp. It would be imprudent to say that you cannot have a fine print without these qualities, but if you choose the optimum solution, you are less likely to be caught out by the chosen subject matter or final image application. Many excellent images have been and will be made with less than ideal equipment and materials. Whether or not they are considered as fine art is another question, which only time will answer.

Measuring Digital Resolution

A direct comparison between digital and analog sources is not easy, since digital cameras, scanners and film systems use different performance measures to evaluate their resolution. It requires some analysis to establish a reliable correlation between the measurement systems, since we use lp/mm to measure film resolution, samples per inch (spi) to measure scanner resolution and sometimes other measures for digital cameras. Scanner specifications themselves can be misleading in two respects. Firstly, by using ‘dpi’ rather than ‘spi’, because dpi, or dots per inch, refers to the resolution of the printed file, whereas spi, or samples per inch, refers to the resolution of the scanning system. Secondly, the equipment spi rating is calculated from the sensor pitch and tracking increment, and takes no account of the effective image resolution of the optical system, which is often found to be lacking. Many scanners are not able to retrieve the full potential of a negative, and one must always evaluate the combined performance of the film system and scanner. One should first consider how digital sensors resolve an image, both theoretically and practically, before setting out to measure their performance. We know that typical digital sensors are made of a regular array of photosensitive elements, and one may wrongly assume that only two lines of sensors are required to resolve a resolution test chart line-pair image, suggesting that 51 spi can distinguish 1 lp/mm. This is a special case, referred to as the Nyquist frequency or cutoff. Although apparently correct, any misalignment between the sensor and the incident image reduces the detected line-pair contrast and, in some conditions, lines are not detected at all. Fig.3 shows how this might happen with the familiar 3-bar pattern of a USAF/1951 chart, imaged onto the array of a typical digital camera sensor. In the first case, where there is almost perfect alignment (fig.3a), the test pattern is fully resolved, but if the image is shifted slightly (fig.3b), no or all pixels are recorded, as all sensor elements see the same image intensity.

With a few calculations, the MTF values for different sensor pitches and alignments can be easily approximated. Reducing the sensor pitch to 2.1 pixels per line pair guarantees line resolution with a minimum of 10% contrast between the lines, and reducing the sensor pitch to 2.6 pixels per line pair guarantees the same with a minimum of 50% contrast. In other words, in order to resolve 1 lp/mm, we need 25.4 × 2.1 or 53 spi for 10% and 67 spi for 50% MTF.

Unfortunately, we can’t just divide the orthogonal pixel count of a sensor by 2.1 to calculate the actual resolution limit of a digital system. An imaging system’s overall MTF is the product of the individual component MTF performances. For instance, the camera lens, film or sensor, digital negative process, enlarging lens and paper all affect the final outcome. However, it would make it difficult and rather confusing if we accounted for all influences of all contributors in the optical system. It is better to understand the impact of each component individually. Consequently, for the rest of the book, we stick to the theoretical values and assume a sampling rate of 53 spi per 1 lp/mm to calculate the resolution limit at 10% MTF, and work with 67 spi per lp/mm to obtain resolution at acceptable sharpness and 50% MTF.

fig.2       Different scanner sampling rates are needed to satisfy the resolution limits of standard vision (10%) and that required to resolve the same detail with acceptable sharpness (50%). The spi figures shown assure a contrast of 10% and 50% at the required lp/mm.

When one rotates either the target or the sensor, the effective sensor pitch decreases, increasing resolution in that orientation. When the grid or sensor is rotated 45°, the effective pitch between diagonal rows of pixels is reduced to 1.86 pixel/lp (fig.3) giving a small resolution improvement of about 13% above the theoretical value. This means that any sensor array, or scanner image, has a range of performance values, depending on the angle of the image. In practice, fine detail like fabric, hair, and grass are oriented at many angles. We assume the worst case orthogonal requirement, since other orientations of this admittedly theoretical image onto a sensor array can produce distracting results, as the image slips between adjacent sensor elements along its length. We shall compare digital camera and scanner resolutions with both orthogonal and diagonal images and use the 53 spi per lp/mm conversion factor to directly compare orthogonal resolution measures, at 10% MTF.

fig.3       Maximum digital resolution changes with alignment and rotation between image sensor and subject detail.

a)           best-case scenario Sensor matrix and test pattern have the same pitch (2 pixels per line pair) and are in almost perfect alignment, which leads to a fully resolved pattern.

b)           worst-case scenario Sensor and test pattern have the same pitch, but the test pattern is moved down by 1/2 pixel. Pixel detection is ambiguous and no pattern is resolved.

c)           improved worst-case scenario Same as top right, but the test pattern pitch is increased to 2.1 pixel/lp. This lowers the actual sensor resolution, but the test pattern is now clearly resolved.

d)           maximum rotation The test pattern is rotated by 45°. This allows the pattern pitch to be reduced to 1.86 pixel/lp. The test pattern is still fully resolved, and the sensor resolution is at its maximum.

MTF as the Standard for Resolution and Sharpness

For this evaluation, we shall use a derivation of the standard Modulation Transfer Function (MTF), described in ‘Sharpness and Depth of Field’. We avoid the traditional fixed test pattern, in three scales and two orientations, in favor of a variable-scale MTF target. This is similar to the Sayce chart, described by Norman Koren (www.normankoren.com), and uses a decreasing pattern scale (increasing lp/mm). The test target is shown in fig.4. The test method requires the test target to be photographed with the camera system set up at a known image magnification. The direct or indirect digital capture method follows and the image is evaluated with imaging software and on a life-size print. We define the system resolution limit as the point where the image contrast reduces to 10% of the maximum contrast possible. The continuous scale in fig.4 is also preferred since it highlights imaging aliasing issues and avoids ‘lucky’ measurements, where perfect alignment of sensor and image occur. At this resolution, the target image becomes a series of faint light and dark mid-gray lines. The limit of acceptable sharpness for the element or system may be implied from the lp/mm value at which the image or print contrast is reduced to 50%. The contrast measurement is accomplished by evaluating the image file or scan and reading the brightness differences between lines with the eyedropper tool of the imaging software.

fig.4       When printed 200mm wide, the scale is exactly 100x larger than life and assumes an image magnification of 1/100. If the image magnification of the test setup is only 1/20, the scale reading should be divided by 5. A copy of this template can be found at www.normankoren.com.

It should be noted that it can be misleading to compare different capture systems’ sharpness, which deploy automatic processing, since sharpness can be radically altered by sharpening algorithms in the imaging software. For instance, it is not uncommon to record a contrast figure exceeding 100% in digital imaging systems, (see fig.8). This is an indicator of over-sharpening of the digital image file, as is a sharp rise in contrast before a dramatic reduction into chaotic oblivion, similar in shape to the filter response in electronics, named after the scientist Chebychev.

Comparing Image Capture Alternatives

The following assessments compare the performance of typical digital SLRs with 35mm roll film and larger film formats, using a range of scanning solutions. The scanning solutions include dedicated film scanners, hybrid flatbed/film scanners and a novel scanning method, which uses a flatbed scanner and a conventional RC print. While the results may change over time, the test methods need not and are described so that one might assess their own equipment and evaluate the latest digital equipment. It is worthy to note that many scanners cannot retrieve the full potential of most lens/film systems, and any increase in available scanner performance will immediately improve your entire image collection, whereas an original digital raw file is as good as it gets.

Dynamic Range

Current digital cameras are not able to record the wide subject brightness range (SBR) that we are accustomed to with monochrome film. Their dynamic range is fundamentally determined by the noise levels of the imaging sensor and its bit-depth. This is easily confirmed by photographing a transmission step tablet placed on a light box. However, the unadjusted dynamic range is already better than slide film, at about 8 stops. Some models claim 9 or 10 stops, but the extra range is not symmetrically distributed about Zone V. In comparison, with appropriate development, a monochrome film can easily record a SBR reaching 15 stops. Typical tonal responses, for slide film, raw digital data, and software-adjusted image data are shown in fig.5. A pictorial comparison, without manipulation, is shown in fig.6a and fig.6b. The dynamic range is established by noting the exposures which produce the digital values of 4 and 96% K, or when printed, reproduce the traditional, absolute reflection densities of 0.09 and 1.89 on the paper.

An analog-to-digital comparison of dynamic range should also consider tonal quality. A digital sensor response is characterized by exaggerated highlight contrast and a long extended shadow toe. To mimic a typical film response, the digital camera exposure must be set so that it does not miss any highlight information (clipped highlights), and it should be recorded in a high-bit file format. The full histogram is then manipulated in the imaging software to reduce the local contrast of highlight regions and boost that of the shadow and midtones (fig.5).

Although extra shadow detail can be recovered by extensive tone-curve manipulation, this action will also accentuate sensor noise, or worse still, if the image is in an 8-bit mode, may cause image tone break-up and posterization in areas of smooth tone. This is a rescue technique, which accentuates sensor noise and does little to tame the highlight appearance. This issue will be overcome as sensors improve their signal to noise ratio (SNR) and their dynamic range is expanded with improvements in analog to digital converter (ADC) resolution.

Throughout this book, the importance of ensuring sufficient negative shadow detail has been emphasized. Conversely, with positive film or digital camera files, the opposite is true, and overexposure is to be avoided at all costs, since it is all too easy to exceed the exposure cut-off point and irretrievably lose highlights. Although some cameras deploy two sensors in each position to improve the highlight response, even these do not appreciably extend the dynamic range. In some cases, slide-film techniques, such as using a graduated neutral density filter to lower a sky’s intensity may avoid subsequent rescue techniques. For static subjects with a large SBR, another technique, called High Dynamic Range, or HDR, using two or more different combined exposures can, with care, capture a full range of subject tones. This technique requires a stationary subject, a tripod and subsequent manipulation. In comparison, the combination of monochrome film and a film scanner effortlessly capture sufficient dynamic range in the most demanding of situations.

fig.5       This graph compares the tonal response of slide film with unadjusted digital raw data and soft-ware adjusted image data. Notice the difference in tonality between the exposure extremes, and see how the slide film quickly rolls off at the exposure extremes while accentuating midtone contrast. Negative film (not shown), normally developed, easily captures the full 10-stop range of this test target.

fig.6       (a) This image is a straight print from a digital SLR, using its raw file setting. Although there is shadow detail, it is tonally compressed and at the same time, the windows are burnt out. To mimic the traditional print, the image would have to be deliberately underexposed and then a correction curve applied to the mid and shadow tones to lift the detail. This accentuates sensor noise.(b) The same scene in a straight print from an Ilford Delta 100 negative, given normal development. There is plenty of detail in the shadows, and on the print, faint tracery is seen in the windows, which can be emphasized by burning-in. Both images were made at an identical ISO and camera exposure setting.

fig.7       This table compares the on-sensor or on-film resolution requirements with the typically measured system performance for a range of formats necessary to deliver sufficient resolution in the final image and satisfy standard and critical observation.

Resolution

After extensive testing and considering our resolution requirements, fig.7 tabulates the required and measured lp/mm and peak imaging capabilities for several sensor and film formats. Fig.9 shows this in graphical form, along with the typical film resolutions after scanning, in relation to the diffraction limit and resolution requirements for several formats. These resolutions were obtained using our MTF test target at the point where the digital image contrast dropped to 10%. These results are discussed later in more detail for each of the different capture solutions.

Images taken with current digital cameras (FX or DX format) have about 1/2 of the effective print resolution of fine-grain, 35mm monochrome film, taken with quality equipment. Their limited pixel count limits their performance. Even as pixel count increases, one should note the required resolution demands placed upon SLR optics from the small DX format are difficult to achieve in practice and, before long, the lens, sensor SNR performance and size will limit the ultimate performance. Assuming an otherwise perfect optical and mechanical system, the maximum print size from a digital camera is calculated by dividing the pixel dimensions by the print ppi setting. An 8x10-inch print, at its standard viewing distance, requires 5.3 lp/mm print resolution, obtained by a 280 ppi file setting (5.3*25.4*2.1 = 280). A 10-megapixel camera meets this requirement, based on the assumption that the image is not cropped and the lens performance comfortably exceeds the sensor resolution. Theory suggests that this sensor should resolve 78 lp/mm orthogonally at 10% MTF contrast but, in practice, it achieves a sensor resolution of only 68 lp/mm, a 12% deterioration. A print crop, changing the print shape from 2:3 to 4:5, further reduces the resolution and takes it below the threshold for a fine print.

As the megapixel race continues, remember that doubling the pixel count only increases resolution by 41%. So, in the case above, it is predictable that a 20-megapixel camera will resolve a minimum of 7.5 lp/mm on a full 8x10-inch print, roughly half that required for critical observation. 35mm film is able to provide up to 11 lp/mm at this enlargement.

Sharpness and Grain

All digital images require some degree of sharpening, either in the capture hardware or in the imaging software, and often, the sharpening occurs behind the scenes. In our comparisons, the images have been optimally sharpened to maximize resolution. In relation to their resolution performance, digital images have sharper images than darkroom prints from negatives. Using the 50% MTF contrast as a guide, the measured results in fig.8 show a gentle softening of contrast for unsharpened digital images and a more abrupt fall-off for sharpened images. Digital images from scanned film can be sharpened too but to a slightly lesser extent than one from a digital SLR. The higher image noise of the scanned image is emphasized by the unsharp mask and drops the optimal sharpening setting to a lower level.

One advantage of digital SLRs is their ability to take pictures at different ISO settings. At similar speed settings, digital SLRs produce smoother images than their 35mm film counterparts, and professional models can challenge medium-format roll film. At the highest ISO settings, however, the smooth imagery that characterizes digital images yields to objectionable noise, which is far less appealing than simple monochromatic high-speed film grain. High ISO settings are best avoided for fine art work, as the noise appearance is intolerable, the dynamic range is reduced and resolution is degraded.

Scanner Assessments

Film Scanner

Darkroom prints repeatedly demonstrate that monochrome film has the potential for sufficient resolution and dynamic range to make fine prints. To capture a negative for the digital domain, one needs some method of digitizing the film. Those models targeted at transparency scanning are advertised on their maximum film density and resolution. Luckily, negative film, in all but extreme circumstances, has a maximum transmission density of 2.0 and is well within the capabilities of all scanners. In this case, the film and development choices are the only limitations for the captured dynamic range. Most dedicated film scanners are able to capture sufficient resolution for a standard-quality print output, even from small format negatives. For example, a high-quality 35mm film scanner will resolve up to 62 lp/mm or 3,300 spi orthogonally, well above half its theoretical 4,800 spi specification. This scanning performance meets the resolution requirement for a full-frame 35mm negative, providing the negative itself has the required resolution, but film grain is obvious at this resolution and degree of enlargement.

fig.8       This chart compares the contrast roll-off for a digital SLR image, at three sharpening levels. The resolution is normalized for the unsharpened digital image. The over-sharpened image shows significantly better sharpness at 50% contrast but ultimately resolves less fine detail, whereas optimum sharpening increases sharpness and resolution from that of the unsharpened image file.

Large and medium-format scanners were initially designed and priced for professional and commercial work, which has since moved over to a full digital workflow. The best medium-format scanners, we have tried, were limited to 56 lp/mm. This is fully sufficient for medium-format but borderline for 35mm negative scanning. A typical large-format scanners will resolve up to 32 lp/mm and, consequently, capture everything a large-format negative has to offer. As more users swapped film for digital cameras, the need and requirements for scanners changed, and excellent models have become harder to obtain.

A high-resolution film scanner will detect and sometimes emphasize film grain. Careful adjustment of the scanning parameters can prevent grain becoming obtrusive, normally accompanied by a small loss in image sharpness. Another consideration is their speed and convenience. While it is convenient to have 24/7 access to a film scanner, the time required to properly scan a negative should not be underestimated. It takes about 20 minutes to clean, preview, adjust, focus and scan a negative properly at a high resolution. This becomes tedious when multiple images require capture. Medium-format and large-format scanners are still specialist items and will remain expensive, to the extent that many photographers consider hybrid flatbed scanners an attractive alternative.

fig.9       Comparing the actual performance of several camera and scanner systems clearly illustrates the predictable quality differences. It clearly shows the barely acceptable performance of the DX format and the increasing performance headroom with increasing format size.

Hybrid Flatbed Film Scanner

Every scanner manufacturer now has a hybrid version with a transparency scanning capability. In the beginning, scanning specifications were poor, at around 300 spi, but have improved dramatically, although the specifications of recent models, exceeding those of dedicated film scanners, are not met in practice. There are two main types: those in which the film is placed into a holder and slid into the body of the scanner, and those that scan the film placed on or close to the glass top, using a second light source in the scanner lid. Theory proposes the former solution to be optimum, since there is no glass plate to distort the optical path and, more importantly, no additional dust attracting surfaces to mar the result. The latter, however, is more popular for cost reasons. The better models hold the film away from the glass surface in special film holders, which prevent the appearance of Newton’s rings in the scanned file and additionally ensure that any dust on the glass is not in the plane of focus.

In practice, the actual performance of flatbed scanners falls short of their quoted specification: For instance, a ‘professional’ flatbed film scanner with a declared resolution of 1,200 × 2,400 spi resolved only 15 lp/mm (800 spi), whereas a later model, claiming to capture 4,800 spi, actually resolved 36 lp/mm (1,900 spi). Both of these scanners fall short of the resolution requirement for detailed images from 35mm, but they provide sufficient resolution for scanning medium or large-format negatives. Hybrid scanners do, however, offer speed advantages over film scanners as a result of less data transfer, shorter lamp warm-up times and the use of a fixed-focus CCD position. It is clear that their optical performance is not as good as their actual CCD resolution, partly due to the fixed-focus design and poor manufacturing tolerances. On consumer hybrid scanners, the optimum plane of focus is frequently not at the position required by the film holder thickness and cannot be adjusted. In these cases, it is worthwhile to experiment with different focus positions by altering the film height with a modified or substitute film holder.

Flatbed Print Scanner

There is a quirky fourth alternative that will produce excellent monochrome scans. This low-budget technique successfully challenges many film scanning solutions at a fraction of the cost, and delivers a significantly higher resolution, by performing the digital capture in two steps. The first step is to make a low contrast enlargement of the image area onto a sheet of 8x10-inch glossy RC paper. The print should be as sharp as possible, using the optimum aperture of the enlarging lens, focused accurately and with the appropriate precautions to minimize enlarger vibration and film waviness. The print should show all shadow and highlight details, so that they can be enhanced or suppressed during digital tonal manipulation.

The second step is to scan the print on a general-purpose flatbed scanner, with a resolution of between 800 and 1,200 spi and preferably using 16-bit depth. The same scanner that resolves 15 lp/mm directly from film can resolve 52 lp/mm through an 8x enlargement of the same 35mm negative and is, consequently, capable of sufficient image resolution.

This method is also able to recover information from overdeveloped or extreme range negatives by scanning two silver prints, made at different print exposure and contrast settings, optimized for either shadow or highlight areas. The two scans may be combined in the photo editing software. With practice, if a print easel is used to accurately locate the prints and the print boundaries are butted against the scanner window edge, the two scanned images will superimpose exactly. These two images can be overlaid and blended together in the photo editing software in a manner analogous to split-grade printing, but with the same computer workload associated with HDR digital manipulation.

This approach produces very good results with inexpensive scanning equipment. It does, however, require extra time to make the initial RC print and so cannot be considered quick. Making an RC print does offer another advantage. It provides a good reference in its own right and can be used to plan the final image manipulations.

General Scanner Performance

Fig.11 compares measured scanning resolutions for several scanner systems and techniques. No scanner is able to capture the full film resolution, and in some cases, they barely meet the minimum requirements of the film formats they were made for, but even the most basic scanner is able to retrieve sufficient resolution from large-format negatives. Unfortunately, we have to assume that the dominance of digital camera sales will ultimately have a detrimental effect on scanner development and model release.

Film Choices for Scanning

You can also consider color film as an image source for monochrome digital prints. This has the advantage that color images can be manipulated and converted to monochrome in new creative ways, opening avenues for self-expression in a monochrome print. An example is shown in the chapter ‘MonoLog’. However, before losing oneself in unbounded digital creativity, one should check that this flexibility is not accompanied by resolution, grain and color sensitivity issues. Color transparency film is not an ideal image source for scanning, because transparencies have a restricted subject brightness range (not unlike digital cameras) and an extremely high density range, which makes them demanding to scan.

In the previous edition, we compared the scanning properties of three emulsion types, Ilford Delta 100, XP2 and Fuji Reala, for resolution, grain and tonality. In practice, we found little resolution difference between the 15x enlargements. Overall, the finest resolution was achieved with fine-grain traditional monochrome film. Conversely, when we compared scanned film grain, those from C41 materials were less obtrusive with a softer grain pattern. Taking into consideration the additional flexibility of color negative originals, the assumption that a scanned monochrome negative is the prime choice for monochrome digital imaging is challenged. In practice, any difference in the color response or tonality between films can be equalized using software adjustments in the photo editing software, either on the monochrome image for overall tonality changes, or on the color image, prior to monochrome conversion, to alter the color sensitivity or mimic the effect of on-camera filters.

fig.10     From the top, these are examples of a dedicated 35mm and a medium-format film scanner (Nikon), a compact large-format film scanner and a hybrid flatbed scanner (Epson).

fig.11     Actual scanner resolutions usually fall short of advertised sampling rates. However, indirectly scanning a print enlargement retrieves far more negative resolution than any direct film scan. For this comparison, each scan was optimally sharpened to maximize image clarity, and the orthogonal resolution was measured at 10% MTF. A subjectively measured extinction resolution, especially along the diagonal axis, is likely to be higher by up to 25%.

Comparing Final Print Quality

Let’s compare analog and digital capture alternatives, using a typical scene (fig.12) with a wide range of detailed textures, smooth tones and man-made objects. A series of photographs were made from the same position with fine-grain monochrome film in a 35mm camera, a medium-format camera and a large-format 4x5 field camera, followed by a digital SLR, each using an equivalent lens at its optimum aperture. The three negatives were scanned, using a hybrid flatbed scanner (4,800 spi), a film scanner or both, and all subsequent image files were printed with the same inkjet printer to create 16x20-inch prints. In addition, a traditional 8x10 and a 16x20-inch darkroom enlargement were made from the 35mm negative. The 8x10-inch enlargement was scanned at 1,200 spi and also printed at 16x20 inches. The 16x20-inch darkroom print was made for comparison purposes. The prints can now be evaluated for fine detail (leaves and tall grass), sharpness (pylon) and grain in smooth tones (sky).

A close examination of the print from the 35mm negative hybrid-flatbed scan (fig.12a) clearly shows that the performance of this capture combination is not adequate to make a detailed print from this film format. A significant improvement using the same negative is made by making an 8x10-inch darkroom print of it first and then scanning it in with a flatbed scanner (fig.12c). Further improvements are seen in the full-scale darkroom print (fig.12d). The dedicated film scanner produces an image of high sharpness and overall contrast (fig.12b), but a close inspection reveals more obvious grain and marginally less detail than the full-scale darkroom print in fig.12d.

A quantum leap in final image quality is seen when using medium-format negative scans (fig.12e&f). Although the film scan (fig.12f) is clearly better than the flatbed scan (fig.12e), both capture rich detail and texture without showing obtrusive grain. Indeed, the film scanner retrieves a level of detail that almost equals the print made from the 4x5 negative scan (fig.12h), which otherwise outperforms all other capture alternatives in this comparison. The print from the digital SLR (fig.12g) has poor resolution, limited by the sensor performance, and it falls behind the best 35mm results, as seen in the blur of leaves and grass. It is, however, virtually grainless at its low ISO setting, and to obtain similar clarity with film, medium or large-format negatives are required.

In conclusion, film is a proven technology and the best method to archive precious images. Film is also a very flexible medium, because it can be printed both traditionally and digitally after scanning. The use of medium or large-format film produces grain-free prints with excellent resolution when using the latest high-resolution hybrid flatbed scanners or dedicated film scanners. A print from a high-resolution scan can reach and exceed the quality of a darkroom enlargement, especially after selective grain reduction and sharpening in the digital domain. As technology continues to improve, digital camera images will increasingly challenge film performance. The flip side is that the same advance in technology also increases digital redundancy and backward compatibilities.

fig.12     This scene was used to compare the relative performance of alternative imaging paths from analog and digital sources. The highlighted area has fine detail, smooth tones and simple well defined structures, which serve to compare the resolution, grain and sharpness in the enlarged samples overleaf.

fig.12a    35mm negative,
flatbed scanner

fig.12b    35mm negative,
35mm film scanner

fig.12c    35mm negative,
8x10 enlargement,
flatbed scanner

fig.12d    35mm negative,
16x20 enlargement

fig.12e    6x7cm negative,
flatbed scanner

fig.12f     6x7cm negative,
medium-format film scanner

fig.12g    10-Mpixel (DX) digital SLR

fig.12h    4x5-inch negative,
flatbed scanner or
large-format film scanner