Adaptive sampling based on fuzzy inference

Monte Carlo is the only choice for a physically correct method to do global illumination in the field of realistic image rendering. Adaptive sampling is an appealing means to reduce noise, which is resulted from the general Monte Carlo global illumination algorithms. In this paper, we take advantage of fuzzy rule-based reasoning to achieve different refinement thresholds for different pixels in the synthesized image. The developed technique can do adaptive sampling elaborately and effectively. Extensive implementation results indicate that our novel method can achieve significantly better than classic ones. To our knowledge, this is the first application of the fuzzy inference to global illumination.