HYBRID OF MEDIAN FILTERING AND AVERAGE FILTERING FOR MONTE CARLO GLOBAL ILLUMINATION

摘要

Monte Carlo methods are relatively simple and are able to solve the global illumination problem for very general scenes. But Monte Carlo algorithms will give very noise images. Besides the costly method such as multiple sampling, nonlinear filtering is a good way to reduce the noise. Since median filter and average filter are simple and are used in many cases, we present a technique using the hybrid of the median filter and the average filter to concentrate our filtering method on pure Monte Carlo global illumination algorithms, such as Monte Carlo path tracing, in order to lower their noise level and improve their visual quality. Computational results show that the filerting method used in this paper is promising.