Efficient occlusion culling for z-buffer systems

Abstract

In recent years the performance of z-buffer hardware has improved rapidly, enabling inexpensive accelerators to smoothly animate relatively simple scenes such as game environments. Although faster processors, buses, and rasterization hardware will gradually enable animation of increasingly complex scenes, it will be many years before very deeply occluded scenes can be rendered in real time with traditional z-buffer hardware.One approach to accelerating the rendering of complex scenes is occlusion culling, which refers to culling of occluded geometry early in the graphics pipeline in order to reduce the workload on transformation, rasterization, and shading hardware. In this talk I will discuss the general problem of occlusion culling, review my work on applying hierarchical methods to the problem, and then describe how these methods can be applied to accelerate hardware pipelines.