自适应初始光子收集半径的卡方渐进光子映射

Chi-square progressive photon mapping based on adaptive initial bandwidth

卡方渐进光子映射(chi-squared progressive photon mapping, CPPM)使用K近邻法(K nearest neighbor, KNN)为命中点确定初始半径,导致图像中光照均匀区域的噪点及明暗交界区域和焦散区域的模糊。针对此问题,提出一种计算光子映射初始半径的算法,自适应地为各命中点确定初始半径:为CPPM算法增加预处理环节,根据对光子分布的均匀程度的检验及对高频区域的筛选为命中点设置初始半径,以保证光照均匀区域的命中点保持在大半径上,光照变化区域的命中点半径快速下降。实验结果表明,改进算法减少了光子映射算法的方差和偏差,提高了渲染效果。

Chi-squared progressive photon mapping(CPPM)uses K nearest neighbor(KNN)to determine the initial bandwidth of the hit point,which leads to the blur of noise in the uniform area of illumination and the blur of the junction area and the caustic area.To solve this problem,an algorithm was proposed to calculate the initial bandwidth of photon mapping,in which the initial bandwidth for each hit point was adaptively determined.The pre-processing step was added for CPPM algorithm,and the initial bandwidth of the hit point was set according to the uniformity of the photon distribution and the screening of the high frequency region,to ensure that the hit point in the uniform region of illumination remained at a large bandwidth,and the hit point bandwidth in the region of illumination changes rapidly decreased.Experimental results show that the improved algorithm reduces the variance and deviation of the photon mapping algorithm and improves the rendering effect.