i
i
i
i
i
i
i
i
12.1. Sources of Three-Dimensional Data 533
of computer aided design (CAD), and often emphasize modeling tools that
correspond to actual machining processes, such as cutting, drilling, etc.
Internally, they will have a computational engine that rigorously manipu-
lates the underlying topological boundaries of the objects. For display and
analysis, such modelers have faceters. A faceter is software that turns the
internal model representation into polygons that can then be displayed. For
example, a model that appears as a sphere internally may be represented
by a center point and a radius, and the faceter could turn it into any num-
ber of triangles or quadrilaterals in order to represent it. Sometimes the
best rendering speedup is the simplest: Turning down the visual accuracy
required when the faceter is employed can increase speed and save storage
space by generating fewer polygons.
An important consideration within CAD work is whether the faceter
being used is designed for graphical rendering. For example, there are
faceters for the finite element method (FEM), which aim to split the surface
into nearly equal-area triangles; such tessellations are strong candidates for
consolidation and simplification, as they contain much graphically useless
data, while also providing no vertex normals. Similarly, some faceters
produce sets of triangles that are ideal for creating actual physical objects
using stereolithography, but that lack vertex normals and are often ill suited
for fast graphical display.
Surface-based modelers do not have a built-in concept of solidity; in-
stead, all objects are thought of in terms of their surfaces. Like solid mod-
elers, they may use internal representations and faceters to display objects
such as spline or subdivision surfaces (see Chapter 13). They may also
allow direct manipulation of surfaces, such as adding or deleting polygons
or vertices. This sort of tool can be used to lower the polygon count of a
model.
There are other types of modelers, such as implicit surface (including
“blobby” metaball) creation systems [117], which work with concepts such
as blends, weights, and fields. These modelers can create impressive organic
effects by generating surfaces that are defined by the solution to some
function f(x, y, z) = 0. Polygonalization techniques are then used to
create sets of polygons for display. See Section 13.3.
Data can also be generated from satellite imagery, by laser scanning, air-
borne LIDAR (light detection and ranging) [571], or other three-dimensional
scanners, by various medical scanning devices (in which image slices are
generated and recombined), and by photogrammetry and computational
photography techniques [1048]. Meshes produced are strong candidates for
simplification techniques, as the data is often sampled at regular intervals,
and many samples have a negligible effect on the visual perception of the
data. Some of these techniques generate nothing more than point clouds,
with no inherent connectivity between the points. Reconstructing meshes