Unbiased, Adaptive Stochastic Sampling for Rendering Inhomogeneous Participating Media

摘要

Realistic rendering of participating media is one of the major subjectsrnin computer graphics. Monte Carlo techniques are widelyrnused for realistic rendering because they provide unbiased solutions,rnwhich converge to exact solutions. Methods based on MonternCarlo techniques generate a number of light paths, each of whichrnconsists of a set of randomly selected scattering events. Finding arnnew scattering event requires free path sampling to determine therndistance from the previous scattering event, and is usually a timeconsumingrnprocess for inhomogeneous participating media. To addressrnthis problem, we propose an adaptive and unbiased samplingrntechnique using kd-tree based space partitioning. A key contributionrnof our method is an automatic scheme that partitions the spatialrndomain into sub-spaces (partitions) based on a cost model that evaluatesrnthe expected sampling cost. The magnitude of performancerngain obtained by our method becomes larger for more inhomogeneousrnmedia, and rises to two orders compared to traditional freernpath sampling techniques.