Preprocessing occlusion for real-time selective refinement

TL;DR: A novel framework that integrates the two techniques to optimize rendering of occlusion computation for densely tessellated models and introduces the use of simplification to automatically deduce virtual occluders to pre-compute visibility.

Abstract: Visibility computation and level of detail modeling are two important components of efficient scene rendering algorithms. Both aim to lessen the graphics load by lowering polygon count. This paper presents a novel framework that integrates the two techniques to optimize rendering. To improve the efficiency of occlusion computation for densely tessellated models, we introduce the use of simplification to automatically deduce virtual occluders. This paper illustrates the use of the technique to efficiently preprocess occlusion for outdoor scenes. Using virtual occluders to pre-compute visibility, only visible surfaces may be refined for real-time selective refinement. Compared to selective refinement without incorporating occlusion culling, our implementation of the framework demonstrates significant polygon count reduction and speedup in frame rate.