Efficient Occlusion Culling with Occupancy Proportion

Visibility culling is a fundamental problem in computer graphics. By combining object-space coherence with image-space coherence, this paper presents a novel algorithm for speeding up complex models rendering. The algorithm is based on the new concept of occupancy proportion (OP), which offers a simple, fast and efficient approach to visibility culling. We exploit the hardware accelerator character, and proposed an asynchronous occlusion queries method which significantly alleviates the affection caused by occlusion queries latency. The algorithm utilizes a hierarchy tree structure containing objects in the scene. The method in the paper can be easily integrated into majority visibility culling algorithms. The algorithm benefits for estimating object visibility and improves the efficiency of object visibility determination. Methodology and implementation details are provided, and experimental results demonstrate that the algorithm can obtain higher rendering efficiency than conservative object bounding box occlusion query method. The algorithm is suited for arbitrary structure scenes especially for slender scenes.