Adaptive Direct Illumination Sampling
Equal-time (20 s) comparison of our method and two other sampling techniques in a scene with 4000 lights: (a) uniform sampling
(fast but poor), (b) sampling proportional to unoccluded light contributions (good but slow), (c) our method.
All images are rendered using path tracing with next event estimation, direct illumination sampling is therefore carried out at each path vertex.
We present a new method for sampling direct illumination in scenes containing a large number of light sources. We divide a scene into cells and find a suitable clustering of the light sources for each of the cells. The clusters are then importance sampled based on their estimated contribution. Our solution is unbiased, scales well with the number of light sources and can be easily incorporated into any existing path tracing rendering system.
Petr Vévoda and Jaroslav Křivánek. 2016. Adaptive direct illumination sampling. In SIGGRAPH ASIA 2016 Posters (SA '16). ACM, New York, NY, USA, Article 43, 2 pages. DOI: https://doi.org/10.1145/3005274.3005283
Links and Downloads
The work was supported by the Charles University Grant Agency, project GA UK 1172416, by the grant SVV-2016-260332, and by the Czech Science Foundation grant 16-18964S. Many thanks to Ludvík Koutný (rawalanche) for modeling the test scene.