蒙特卡罗与准蒙特卡罗相互融合的整体光照计算

蒙特卡罗方法具备普适性、鲁棒性以及与问题复杂度无关性等优点，非常适于十分难解的整体光照计算问题，但缺点是生成图像随机噪声大．准蒙特卡罗方法计算连续被积函数低维积分的收敛速度快于蒙特卡罗方法，但不适于直接求解复杂的整体光照计算问题．文中研究蒙特卡罗整体光照计算最根本环节，即随机游动的抽样模式，提出融合蒙特卡罗与准蒙特卡罗的两种通用的新型整体光照计算策略．两种新型策略可以应用于所有基于蒙特卡罗的整体光照算法，不仅能够降低生成图像的随机噪声，而且实现简单、不增加计算和存储开销．

Fusion of Monte Carlo and Quasi-Monte Carlo for Global Illumination

Monte Carlo method performs well in the computation of complex global illumination problem because it is general, robust and independent of dimension. However, the synthetic image produced by Monte Carlo is too noisy. Low dimensional quasi-Monte Carlo integration for continuous integrand converges faster than Monte Carlo, but quasi-Monte Carlo is not fit for complex global illumination when it is used directly. In this paper, we study the sampling procedure for the random walk, which is the fundamental of Monte Carlo global illumination. Based on the fusion of Monte Carlo and quasi-Monte Carlo, we present two novel computational strategies for solving the global illumination problem. The two strategies are general and can be used for all of the Monte Carlo based global illumination algorithms. The proposed strategies can not only lead to a reduction of image noises, but it could be also implemented easily without increasing computation and memory costs.