Tighter Bounding Volumes for Better Occlusion Culling Performance

Bounding volumes are used in computer graphics to approximate the actual geometric shape of an object in a scene. The main intention is to reduce the costs associated with visibility or interference tests. The bounding volumes most commonly used have been axis-aligned bounding boxes and bounding spheres. In this paper, we propose the use of discrete orientation polytopes (k-dops) as bounding volumes for the specific use of visibility culling. Occlusion tests are computed more accurately using k-dops, but most importantly, they are also computed more efficiently. We illustrate this point through a series of experiments using a wide range of data models under varying viewing conditions. Although no bounding volume works the best in every situation, k-dops are often the best, and also work very well in those cases where they are not the best, therefore they provide good results without having to analyze applications and different bounding volumes.