Global Illumination Noise Analysis And User-guided Rendering

Monte Carlo ray tracing is the fundamental algorithm in the global illumination area. Its accuracy and flexibility have been recognized by the computer graphics community. At the meantime, it takes many computation resource to produce photo-realistic synthesize image due to its slow convergence. To make the algorithm more efficient is still one of the most important issues in this area.In this paper, we analyze the noise produced by Monte Carlo ray tracing and also describe a new method called user-guided bi-directional path tracing. First, we introduce the Monte Carlo method and its noise. To reduce the noise, we analyze the noise produced by sampling, scattering and transportation. We illustrate its distribution, main trend and deviation in its convergence. Some possible schemes in the future research are also discussed. The user-guided bi-directional path tracing is proposed as the fact that the origin one is not configurable and also can't manipulate some difficulties, which are easy to deal with by humans. According to the principle of importance sampling, the new method can keep track on the characteristic of the scene being rendered at very low cost. It not only enhances the efficiency but also the flexibility and controllability. At the end, we describe the alteration we made to the PBRT rendering system and briefly introduce the bi-directional path tracing implementation.