Dynamic scene occlusion culling

摘要

Large, complex 3D scenes are best rendered in an output-sensitivenway, i.e., in time largely independent of the entire scene model'sncomplexity. Occlusion culling is one of the key techniques fornoutput-sensitive rendering. We generalize existing occlusion cullingnalgorithms, intended for static scenes, to handle dynamic scenes havingnnumerous moving objects. The data structure used by an occlusion cullingnmethod is updated to reflect the objects' possible positions. To avoidnupdating the structure for every dynamic object at each frame, antemporal bounding volume (TBV) is created for each occluded dynamicnobject, using some known constraints on the object's motion. The TBV isninserted into the structure instead of the object. Subsequently, thenobject is ignored as long as the TBV is occluded and guaranteed toncontain the object. The generalized algorithms' rendering time isnlinearly affected only by the scene's visible parts, not by hidden partsnor by occluded dynamic objects. Our techniques also save communicationsnin distributed graphic systems, e.g., multiuser virtual environments, byneliminating update messages for hidden dynamic objects. We demonstratenthe adaptation of two occlusion culling algorithms to dynamic scenes:nhierarchical Z-buffering and BSP tree projection