Accelerating Ray Shooting through Aggressive 5D Visibility Pre-processing

We present a new approach to accelerating general ray shooting. Our technique uses a five-dimensional ray space partition and is based on the classic ray-classification algorithm. Where the original algorithm evaluates intersection candidates at run-time, our solution evaluates them as a preprocess.

The offline nature of our solution allows for an adaptive subdivision of ray space. The advantage being, that it allows for the placement of a user set upper bound on the number of primitives intersected.

The candidate sets produced account for occlusion, thereby reducing memory requirements and accellerating the ray shooting process. A novel algorithm which exploits graphics hardware is used to evaluate the candidate sets. It is the treatment of occlussion that allows for the practical precomputation of the ray space partition. This algorithm is called aggressive since it is optimal (no visible primitives are included), but may result in false exclusion of visible primitives. Error is minimised through the adaptive sampling.