In [109], Zhang suggests building a depth estimation buffer by computing the farthest depth value in the projected bounding box for an occluder and using this value throughout the screen-space bounding rectangle for the occluder. The end result is a tiling of the image plane with a set of projected occluders each representing a single depth value, as shown in Figure 26. The computation is kept simple to avoid complex scan-conversion of the occluder and to simplify the depth comparisons against a candidate occluded object.

2001-01-10