The Study Of Path Tracing Based Energy Redistribution Global Illumination Method

A new global illumination algorithm can be got by combining Monte Carlo path tracing with Metropolis Light Transport mutation strategies: Energy Redistribution Algorithm. It's used to solve correlated integral problems. The algorithm works by redistributing the energy of initial path traced samples over the image plane. The heart of ER algorithm lies in choosing a flow filter to redistribute the energy of the initial MC samples.The main content of this thesis is to design the energy flow filter in energy redistribution sampling method, to realize the energy redistribution light tracer based on path tracing and to apply it in the field of animation. At first, the Monte Carlo method is introduced. And then, the basic energy redistribution sampling method is introduced. Later, two energy flow filters named equal deposition flow filter and energy dependent equal deposition flow filter are proposed. At last, the energy redistribution path tracing algorithm is got by combining energy redistribution sampling and path tracing. First some seed paths are generated at each pixel use path tracing, and from each seed path, a sequence of paths are generated by mutating the current path using the mutation strategies. Each mutation is accepted or rejected with calculated acceptance probability, and given some energy of the initial seed path to them. As the process continued, the energy distribution converges to a stationary distribution the same as the image contribution. Three path mutation strategies are introduced meanwhile. The details of the energy redistribution path tracer are also included. At last, the application of energy redistribution path tracing in animation is introduced in detail. First the alteration of the render system is introduced, then, the paths of the scene are classified and the reuse strategies of the different classes of these paths are introduced: if the seed path can be reused, the same seed path is recorded in other frames; otherwise a new seed path is sampled. And then, if the path it got can be reused, the acceptance probability must be calculated to decide whether accept it or not; otherwise, the current path recorded in other frame is mutated to get a new path and the acceptance probability is calculated to decide whether accept it or not .At last some results of the animation with object moving and with camera changing are exhibited.