A Cost Metric for Scene-Interior Ray Origins

Acceleration structures reducing the number of intersection tests are crucial for high ray tracing performance. The best acceleration structures are currently obtained using the surface area heuristic (SAH) which greedily optimizes a cost metric based on a series of heuristic assumptions about the rays. Noting that rays in many cases originate within the scene’s bounding box, we assume a uniform distribution of their origins throughout the scene and derive alternative cost metrics. Our experimental results show that at a slight increase in acceleration structure build time, an improvement in ray tracing speed over the SAH is achieved for most scenes. A more accurate metric based on our assumptions yields even more reliable speed-ups at the cost of higher construction time. We conclude that making more realistic assumptions than the SAH is a promising route for better ray tracing acceleration.

[1]  K. Joy,et al.  A hybrid CPU-GPU Implementation for Interactive Ray-Tracing of Dynamic Scenes , 2008 .

[2]  W.A. Hunt Corrections to the surface area metric with respect to mail-boxing , 2008, 2008 IEEE Symposium on Interactive Ray Tracing.

[3]  I. Wald,et al.  On building fast kd-Trees for Ray Tracing, and on doing that in O(N log N) , 2006, 2006 IEEE Symposium on Interactive Ray Tracing.

[4]  W.R. Mark,et al.  Fast kd-tree Construction with an Adaptive Error-Bounded Heuristic , 2006, 2006 IEEE Symposium on Interactive Ray Tracing.

[5]  A. van Oosterom,et al.  The Solid Angle of a Plane Triangle , 1983, IEEE Transactions on Biomedical Engineering.

[6]  Alexander Reshetov,et al.  Fast Ray Tracing for Modern General Purpose CPU , 2002 .

[7]  Chris Fowler,et al.  Implementing the RT2 Real-Time Ray-Tracing System , 2007 .

[8]  S. Popov,et al.  Experiences with Streaming Construction of SAH KD-Trees , 2006, 2006 IEEE Symposium on Interactive Ray Tracing.

[9]  Ingo Wald,et al.  Realtime ray tracing and interactive global illumination , 2004, Ausgezeichnete Informatikdissertationen.

[10]  Vlastimil Havran,et al.  On Improving KD-Trees for Ray Shooting , 2002 .

[11]  Kellogg S. Booth,et al.  Heuristics for ray tracing using space subdivision , 1990, The Visual Computer.