A scene with complex occlusion rendered with depth of field. Left: Images rendered by PBRT [Pharr and Humphreys 2010] using 16 and 256 low-discrepancy samples per pixel (spp) and traditional axis-aligned filtering. Right: Image reconstructed by our algorithm in 10 seconds from the same 16 samples per pixel. We obtain defocus quality similar to the 256 spp result in approximately 1/16th of the time.
Paper: PDF (SIGGRAPH 2011 preprint)
Videos: ZIP
Slides: Keynote '09, zipped (66 MB)
Slides: PDF (45 MB)
Code: github
Traditionally, effects that require evaluating multidimensional integrals for each pixel, such as motion blur, depth of field, and soft shadows, suffer from noise due to the variance of the highdimensional integrand. In this paper, we describe a general reconstruction technique that exploits the anisotropy in the temporal light field and permits efficient reuse of samples between pixels, multiplying the effective sampling rate by a large factor. We show that our technique can be applied in situations that are challenging or impossible for previous anisotropic reconstruction methods, and that it can yield good results with very sparse inputs. We demonstrate our method for simultaneous motion blur, depth of field, and soft shadows.