基于光线追踪的全局光照及降噪处理研究  

Research on Global Illumination and Denoising Based on Ray Tracing

作　　者：李学阔 温佩贤 杨林 方泽华 杜晓荣 LI Xue-kuo;WEN Pei-xian;YANG Lin;FANG Ze-hua;DU Xiao-rong(School of Physics,Sun Yat-Sen University,Zhuhai 519000,China;Kingsoft Corporation,Zhuhai 519000,China)

机构地区：[1]中山大学物理学院,广东珠海519000 [2]金山网络科技有限公司,广东珠海519000

出　　处：《计算机技术与发展》2022年第9期23-27,共5页Computer Technology and Development

基　　金：广东省重点领域研发计划资助项目(2019B010148001)。

摘　　要：传统的路径追踪算法能够有效地追踪光线的运动轨迹,从而在屏幕上呈现出真实的渲染效果,但传统的光线追踪算法只能够追踪镜面反射或规则的折射光,追踪漫反射光消耗太大,效率低下。基于蒙特卡洛概率方法的光线追踪能够从统计意义上很好地实现物体间漫反射的光照效果,极大地解决了传统光线追踪的缺陷和效率问题,然而只有当采样数为无限大时,蒙特卡洛方法才是无误差的。为了提升渲染性能和质量,尽可能缩小误差,提出了一种根据颜色采样分布进行降噪处理的优化方法,该方法可以减少采样数,通过后期处理来去除噪声。通过在自研图形引擎中渲染虚拟场景对比实验发现,根据颜色采样分布进行降噪处理的方法能够实现较好的渲染及降噪效果,提高整个算法的性能以及画面表现。The traditional path tracing algorithm can effectively track the trajectory of light,thereby presenting the real rendering effect on the screen,but it can only track the specular reflection or regular refracted light,tracking diffuse light is inefficient.Ray tracing based on Monte Carlo probability method can achieve the lighting effect of diffuse reflection between objects in a statistical sense,which greatly solves the defects and efficiency problems of traditional ray tracing.However,Monte Carlo method is error free only when the number of samples is infinite.In order to improve the rendering performance and quality and reduce the error as much as possible,an optimization method for denoising according to the color sampling distribution is proposed.This method can reduce the number of samples and remove the noise through post-processing.Through the contrast experiment of rendering virtual scene in self-developed graphics engine,it is found that the method of denoising according to color sampling distribution can achieve better rendering and denoising effect,and improve the performance of the entire algorithm.

关 键 词：全局光照 光线追踪 蒙特卡洛 采样数 降噪 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]