Developing accurate and efficient cost metrics for constructing, traversing and distributing hierarchical acceleration structures for ray tracing based algorithms.

Rendering is one of the major research topics in the field of computer graphics. It deals with the generation of images depicting a virtual world. Ray tracing based renderers trace rays through a virtual 3D scene, which need to be intersected with this scene’s geometry to resolve visibility and are used in shading calculations. Acceleration data structures exploit spatial coherence by partitioning the scene’s geometry into spatial groups. If a group is not relevant to a given ray, the associated geometric shapes do not need to be tested for intersection. Hierarchical structures achieve the largest reduction in intersection tests, but are non-trivial to construct optimally for given ray distributions. This project aims at developing new accurate cost metrics (i) for deciding on the optimal partitioning of the scene’s geometry for constructing tree acceleration data structures, (ii) for deciding on the optimal order of the nodes to traverse while tracing rays through these tree structures and (iii) for deciding on the optimal distribution of the scene’s geometry over multiple rendering machines. Here optimality is defined in the least number of intersection tests to perform and the least number of rendering machines to visit for tracing a ray, which is crucial for increasing the rendering performance. Furthermore, this project aims at developing new techniques for efficiently evaluating the cost metrics irrespective of the number of geometric shapes constituting the 3D scenes.