Game Development Reference
In this chapter we will look at how to reduce polygons from your mesh. In
essence this is the opposite of the previous chapter. Whereas in the
previous chapter we were concerned with smoothing the display of a low
polygon mesh by using progressive refinement of the mesh, in this
chapter we will look at a technique for reducing the polygons in a mesh.
The aim is to gradually reject polygons until a target polygon total is
achieved. There are many options for polygon reduction and we will look
briefly at these alternatives before examining one technique in detail.
Michael Garland and Paul S. Heckbert first introduced the one method
that I felt gave the best results for low polygon meshes. Their method is
known as Quadric Error Metrics and is discussed in detail in this
Painting polygons is the most time-consuming aspect of generating real-
time computer graphic displays. If instead of painting 10 000 polygons we
can paint just 1000, then the display update time will improve dramatically.
As a character heads off into the distance, the area of the screen display
occupied by this character will be reduced. Suppose your key character
uses 2000 polygons. At some stages in the game her on-screen height is
500 pixels. At another stage her height occupies just 50 pixels. If she
looks good using 2000 polygons when 500 pixels high, then using 2000
polygons for the 50 pixels high version is unnecessarily complex. The
purpose of this chapter is to show how to create several versions of the
character that are suited for use at different distances from the camera.
Most modelling software will include tools for polygon reduction. One
option would be to include the different resolution meshes in the
application. While this approach would be suitable for CD-ROM distribu-
tion, it does not suit low bandwidth Internet applications. For the latter, we