Assessing Alternatives to the Surface Area Heuristic for Bounding Volume Hierarchy Construction

Abstract

In this thesis, alternatives to the greedily applied Surface Area Heuristics for BVH construction for ray tracing are explored. Non-greedy evaluation is used for measuring the impact of greedy evaluation methods and other heuristics are evaluated to see the ray tracing impact separately by ray distribution and scene. Greedy construction for random scenes shows linear growth in impact compared to non-greedy construction, but minimal impact for scenes with more structured primitives. Alternatives to the Surface Area Heuristic itself show improved ray tracing performance in given scenes, but on average, SAH still is found to be one of the better alternatives. It is found that if the heuristic can be fitted to the scene, better alternatives to SAH exist.