Visibility culling for interactive dynamic scenes

Virtual environments are rapidly growing in size and complexity. At the sametime, there is a strong commercial need for rendering larger and larger scenesat interactive rates. This leads to two basic enhancements, one by increasingthe performance and the size of memory in hardware in order to support largescene rendering and another in software by designing more efficient visibilityalgorithms. Visibility culling methods manage complexity by sending only thepotentially visible primitives into the rendering pipeline. At present, occlusionculling algorithms do not handle well scenes with dynamic objects. One of themain difficulties is handling changes to the object hierarchies since the visibilityinformation changes continuously. In this chapter, we present a fast from-regionocclusion culling method that is able to compute the potential visible sets onlinefor large dynamic outdoor scenes. The method uses an occlusion map on a dualray-space in order to encode visibility with respect to a view cell. It utilizes newfeatures of the advanced graphics hardware architecture to construct and maintainocclusion maps.