基于去噪光子图的光子映射

光子映射是产生高质量的照片级图像的有效方法。基于内存中有限的M个光子,提出基于去噪光子图的光子映射,即通过光子能量传递,得到若干个包含M个去噪光子的光子图保存到外存中;在光照估计阶段,将外存中这些去噪光子图依次读入内存,进行光照估计而得到相应的图像,最终的图像为上述图像的平均。这大大减少了光照估计的噪声,通过设置较小的光照估计的最近光子数目,可以得到噪声和偏差都很小的图像,并且视点切换时无需重新进行光子跟踪,即绘制时间非常少。实验表明只需较少的去噪光子图,就能得到很好的光照估计的效果。

Photon Mapping Based on Noise-reduced Photon Maps

Photon mapping is an efficient method which can produce high-quality photorealistic images.Base on M photons in memory,photon mapping was proposed based on noise-reduced photon maps.By energy distribution between photons,some photon maps were saved including M noise-reduced photons into external memory.In photon map estimation stage,these noise-reduced photon maps in external memory were loaded into memory successively,and illumination for each photon map was estimated and corresponding images was got,and final image is the average of these images.This can largely reduce noises of photon maps,so the image is got whose noise and bias are very small by setting small number of the nearest photons of illumination estimation.And switching viewpoint does not need to carry out photon tracing again,so rendering time is very small.The experiments show that fine effect of illumination estimation can be got by a few noise-reduced photon maps.