Pablo Bauszat, Marc A. Kastner, Martin Eisemann, Marcus Magnor
We present a new acceleration structure for ray tracing called the Split Grid. Combining concepts of hierarchical grids, kd-trees and Bounding Volume Hierarchies (BVHs), our approach is based on the idea of nesting 1D-grids. Our proposed acceleration structure is compact in storage, adaptive to the scene geometry and can be traversed using a fast and efficient traversal scheme. We show that the Split Grid is comparable to other current state-of-the-art acceleration structures regarding traversal performance and memory footprint. While other data structures usually achieve these levels of performance only due to a complex and expensive construction process (e.g. using the Surface Area Heuristic), our proposed Split Grid is built with a very simplistic construction scheme which is a major benefit of our approach.
Type: Poster at Eurographics
Date: March 2014