Photon Splatting for participating media

Since the beginning of image synthesis, much research have been done on global illumination simulation. However, simulation in participating media is still an open problem as far as computing time is concerned. Recently some methods, like Photon Mapping, proposed an optimization of the resolution of global illumination in participating media. Nevertheless, the computing costs of these methods remain very expensive.In this paper, we present a method which takes advantage of density estimation to efficiently reconstruct volume irradiance from the photon map. Our idea is to separate the computation of emission, absorption and out-scattering from the computation of in-scattering. Then we use a dual approach of density estimation to optimize this last part as it is the most computational expensive. Our method extends Photon Splatting, which optimizes the computation time of Photon Mapping for surface rendering, to participating media, and then considerably reduce participating media rendering times. Even though our method is faster than Photon Mapping for equal quality, we also propose a GPU based optimization of our algorithm.