Shape Based Culling

This paper describes a technique to improve the rendering time of volumetric datasets using a volumetric skeleton. The skeleton is used to define a hierarchical set of bounding spheres which completely encapsulate the volume, enabling efficient space leaping during the ray casting and ray tracing processes. The spheres can also be used to cull portions of the dataset, i.e. by specifying logical sections of data to remove. This is much more intuitive than trying to accurately place cutting planes which can be cumbersome. We also show how intuitive culling and selection can produce different types of special effects. Since the bounding spheres are computed from the skeleton, and the skeleton is used for animation, the bounding spheres just have to be transformed to enclose a moving volumetric object and do not need to be recomputed. This creates a significant improvement in the rendering times for animating volumes. The hierarchical spheres can also be used for collision detection. This allows a volume graphics system to compute animation, collision detection and volume rendering with one basic data structure.