In the previous chapters, you learned various image processing techniques related to image manipulation. In some of them, we concentrated our processing on specific regions of the images, predefined by the user. However, many processes that involve visual media enhancement need to focus on automatically specified regions of interest. In this chapter, we will present some basic techniques for selecting the regions of interest, based on image morphology. We will also revisit the manual selection of regions, presenting some more flexible tools. Then, you will be demonstrated some basic object analysis techniques such as edge, corner, and circle detection. Several examples will help you better understand how morphological operations combined with object analysis methods can help in targeting our processing on specific areas of an image.
In this chapter, we shall:
- Learn about binary images and how they are used for masking
- Learn about morphological operations and their importance
- Learn how to use MATLAB tools for Region Of Interest (ROI) selection
- Learn how to detect edges, corners, and circles in an image
So, let's start!
To understand the notion of morphological operations, we will have to revisit the thresholding techniques presented in the previous chapter. We have already mentioned that thresholding an image leads to binary images, which are defined by their two possible pixel values; 0 (for black) and 1 (for white). The way to convert a grayscale image to binary is through thresholding; that is, setting the pixels above a certain value to 1 and the rest to 0. Let's now explain the basic reasons for binarizing an image. The purpose of image binarization can be split into two levels. At a first level, it is used to pinpoint the pixels of an image that interest us (usually called regions of interest or simply, ROIs), thus giving us a quick and easy overview of the image content. The binary images derived, are often called masks. At a second level, it can be used for processing only the selected ROIs (with pixel values equal to 1) defined by the mask, leaving the rest of the image unaffected. Let's see the difference using, an example that covers both the functionalities.