Towards Adaptive Occlusion Culling Using Camera Coherence
Ignacio Mansa, Aiert Amundarain, Eleazar Elizalde, A. García‐Alonso, Luis Matey · 2006
Occlusion culling proves to be useful for the interactive visualization of environments that are not densely occluded. Those which are built up by dense geometric sets like aerospace engines composed of thousands of components and millions of polygons. In first place the convenience of using occlusion culling is studied with a simple scheme. Then improvements are analyzed. The key points to obtain frame rate speed-ups are: a convenient occlusion query scheduling provides the performance required; depth sorting is performed only when camera orientation changes more than a given threshold; coherence reduces the number of occlusion queries posted per frame. It is possible to select the percentage of occlusion queries that will be performed in each frame, from non-conservative schemes up to a conservative one. Furthermore, we propose a small addition to the GPU occlusion queries to perform faster renderings