Path Differentials for Monte Carlo Rendering
Frank Suykens |
Contact: Computer Graphics Research Group
Dagstuhl Seminar 01242: Stochastic Methods in Rendering
Wadern, Germany, 10-15 June 2001
Abstract
Photo-realistic rendering algorithms such as Monte Carlo ray tracing sample individual paths to compute images. Noise and aliasing artefacts are usually reduced by supersampling. Knowledge about the neighborhood of the path, such as an estimated footprint, can be used to reduce these artefacts without having to trace additional paths. Ray differentials estimate such a footprint for classical ray tracing, by computing ray derivatives with respect to the image plane. The footprint proves useful for filtering textures locally on surfaces. Path differentials generalize this idea to arbitrary Monte Carlo path sampling, including general reflection and refraction functions. Sampling new directions introduces additional partial derivatives, which can all be combined into a footprint estimate. Additionally the path gradient is introduced; it gives the rate of change of the path contribution. When this change is too steep the size of the footprint is reduced. The resulting footprint can be used in any global illumination algorithm that is based on path sampling. We have developped two applications to show its potential: texture filtering in distributed ray tracing and a new approach to hierarchical refinement in particle tracing radiosity.
Keywords: BRDF sampling, global illumination, Monte Carlo, ray differentials