Archived

This topic is now archived and is closed to further replies.

foobar

3d occlusion algorithms

Recommended Posts

foobar    122
has anyone links to average-level (not totally mathematical, not totally stupid) description(s) of fast and general-case 3d occlusion culling algorithms. though they should discribe rather detailed what to do and what to expect. interesting ones would be: beam-trees anti-portal-BSP (just heard about it, dont know what this is ) occluder-combining/joining and any hirarchical z-buffer stuff. what i basically want to do is: see if either out of AABB, bounding-sphere, view-parallel-circle/rect is partially visible. the simplest way would be to allocate a software-z-buffer, draw every shape with a constant z-value (taken from the farthest point) and to compare the occludee-covered pixels (taking the nearest z-value of the occludee) with the contents of the z-buffer. with "straight" z-buffer this would be slower than rendering then whole invisible geometry twice with any FAST "c-buffer with z-information" this should work, and i THINK that hirarchical z-buffers are kindof. any input appreciated, except on my bad english aaaah: i forgot: i dont want to draw STRICT front-to-back, since i want to do frustrum-culling (at least stage 1) with an "object-based" loose octree, and use the "loose-z-order" i get from the octree to draw "loose front-to-back". thx, bye, Edited by - foobar on January 31, 2002 5:30:37 PM

Share this post


Link to post
Share on other sites