Scalable photon splatting for global illumination

In this paper, we present a new image based method for computing efficiently global illumination using graphics hardware. We propose a two pass method to compute global lighting at each pixel. In the first pass, photons are traced from the light sources and their hit points are stored. Then, in the second pass, each photons hit point is splatted on the image to reconstruct the irradiance. The main advantages of our method in comparison with previous approaches is scalability. Indeed, it can be used to obtain coarse solution as well as accurate solution and is well suited for complex scenes. First, we present the density estimation framework that is the theory behind our reconstruction pass to understand how to control the accuracy with our method. Then, we show how to use graphics hardware to improve reconstruction time with some restrictions but without losing the ability to compute an accurate simulation.