基于推迟重采样的时空路径复用蓄水池算法  

Spatiotemporal path-reusing reservoir algorithm based on deferred resampling

作　　者：刘双嘉 马宁 李方钏 张严辞[1,2] Liu Shuangjia;Ma Ning;Li Fangchuan;Zhang Yanci(College of Computer Science,Sichuan University,Chengdu 610065,China;National Key Laboratory of Fundamental Science on Synthetic Vision,Sichuan University,Chengdu 610065,China)

机构地区：[1]四川大学计算机学院,成都610065 [2]四川大学视觉合成图形图像技术国家级重点实验室,成都610065

出　　处：《计算机应用研究》2024年第12期3843-3850,共8页Application Research of Computers

基　　金：国家重大专项资助项目(GJXM92579);四川省重点研发资助项目(2023YFG0122)。

摘　　要：现有的基于蓄水池的时空重要性重采样(ReSTIR)算法在渲染具有光泽(glossy)材质的场景表面时,难以兼顾渲染质量和性能。为此提出了一种基于推迟重采样的时空路径复用蓄水池算法。其基本思想是仅复用起点为漫反射(diffuse)材质的路径和子路径,利用diffuse采样分布的弱方向性,减少采样分布差异。具体而言,该算法将样本获取和重采样计算,从着色点推迟到路径上第一个diffuse材质的路径顶点。此外,提出了一种路径选择方法,通过选择推迟距离更小的路径,来增强重采样结果对渲染结果的影响。实验结果表明,与其他ReSTIR算法相比,该算法能够以较低的性能开销,取得较高质量的glossy表面渲染结果,在实时渲染中具有更高的实用价值。Existing reservoir-based spatiotemporal importance resampling(ReSTIR)algorithms struggle to balance rendering quality and performance when rendering scenes with glossy materials.To address this issue,this paper proposed a spatiotemporal path-reusing reservoir algorithm based on deferred resampling.The basic idea was to only reuse paths and sub-paths with a diffuse starting point,leveraging the weak directionality of diffuse sample distribution to reduce sample distribution differences.Specifically,this algorithm deferred sampling and resampling from the shading point to the first diffuse vertex on the path.In addition,this paper proposed a path-selecting method to enhance the impact of resampling results on rendering results by selecting paths with smaller deferred distances.The experimental results show that compared to other ReSTIR algorithms,the proposed algorithm can achieve high-quality rendering results of glossy surfaces with low overhead,making it more valuable in real-time rendering.

关 键 词：实时渲染 路径追踪 路径复用 光泽材质 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]