基于模糊不确定性的自适应采样

Adaptive Sampling Based on Fuzzy Uncertainty

提出一种用于Monte Carlo全局光照的自适应采样方法,使得合成图像时对每个像素采用不同的采样数量,以提高间接光照的表现效果,降低图像总体噪声水平.考虑到图像或像素噪声水平的评价具有内在的模糊不确定性,基于模糊理论,以像素样本光照为基本元素建立模糊集合,利用模糊集的模糊度提出一种新的像素噪声水平评价标准.在新评价标准的基础之上实施自适应采样,首先对像素进行少量采样,然后根据新标准评价其噪声水平,并有针对性地对噪声水平较高的像素使用较多的采样样本.通过大量实验,验证了文中方法比已有的自适应采样方法更好.

An adaptive sampling technique to calculate the global illumination with Monte Carlo path tracing is presented. It takes different number of sampling for different pixel to enhance the indirect illumination and to reduce image noise. Based on the intrinsic fuzzy uncertainty in image noise estimation, we define a fuzzy set based on the initial sampling results for each pixel and propose a new noise metric by exploiting the idea of fuzziness defined in fuzzy set theory. With the proposed noise metric, we can perform efficient adaptive sampling to determine whether super sampling is needed or not for each pixel. Extensive experiment results show that our novel method can achieve significantly better results than presently existing algorithms.