Video (showing runtime performance), QuickTime(19MB)
Video (showing image quality), QuickTime(48MB)
(You can download QuickTime from QuickTime)
|
|
| Scan of Michelangelo's St. Matthew: We use our LOD-based algorithm to accelerate ray tracing of the St. Matthew model with shadows and reflections. We ray trace the 128M triangle model at 512x512 resolution and achieve 2-3 frames per second on a dual Xeon processor workstation with 4GB of memory. We observe 2-20 times increase in the frame rate due to R-LODs with very little loss in image quality. | Forest Model: We render the forest model consisting of 32 million triangles with shadow rays. The image resolution is 512 by 512 with 2x2 super-sampling and we set PoE = 4. We are able to render the model given the viewpoint at 1.6 frames per second and achieve 5 times improvement by using R-LODs. |
We present a novel LOD (level-of-detail) algorithm to accelerate ray tracing of large models. Our approach computes drastic simplifications of the model and the LODs are well integrated with the kd-tree data structure. We introduce a simple and efficient LOD metric to bound the error for primary and secondary rays. The LOD representation has small runtime overhead and our algorithm can be combined with spatial coherence techniques and cache-coherent layouts to improve the performance. In practice, the use of LODs can alleviate aliasing artifacts and improve memory coherence. We implement our algorithm on both 32-bit and 64-bit machines and are able to achieve up to 2-20 times improvement in frame rate of rendering models consisting of tens or hundreds of millions of triangle with little loss in image quality.
|
 |
||
| The use of R-LODs can alleviate aliasing artifacts. The image on the right is ray traced without LODs and has aliasing artifacts. The image on the left is generated with PoE = 4 and has fewer aliasing artifacts | Double Eagle Tanker: The deck of the Double Eagle tanker with shadows is shown using ray tracing. We are able to achieve 1-3fps at 512 by 512 image resolution with 2x2 super-sampling on a dual Xeon workstation. In this model, the working set of the ray tracer is low and we are able to achieve up to 2 times improvement in the frame rate. |
Cache-Oblivious Layouts of Bounding Volume Hierarchies
CB #3175,
Department of Computer Science
University of North Carolina
Chapel Hill, NC 27599-3175
