Research In Metropolis Light Transport Based On Simulated Tempering

Computer graphics plays a very important role in computer application.There is a basic and crucial problem that exists in computer graphics.That is to convert a scene to a picture,which is named rendering.As the theory develops,rendering has changed from an experimental model to a physically based model derived from a simplified Maxwell equations.The most important achievement was the rendering equation,which gave photo-realistic rendering a acurate mathamatical model.Rendering equation is a integral equation that is hard to solve by regular methods.Fortunately,researchers have conducted a pure integral from the rendering equation.It is named path integral and lots of global illumination algorithm have been developed basing on it.One of these algorithms is multiplexed Metropolis light transport.The algorithm samples by Metropolis adaptive sampling combining with simulated tempering that can sample a series of functions at the same time,and then estimates the integral by the samples.This paper analyzes on multiplexed Metropolis light transport,points out its pros and cons.By targeting its shortcoming on re-using sub-path and introducing some ideas such as virtual point light and virtual point sensor from bidirectional lightcuts,we design a new algorithm that has improvement compared with the original algorithm in aspect such as sub-path re-using,pathspace continuity,zero-value region and so on.We also design a weighting strategy of multiple importance sampling that is suitable for our algorithm,to better combine the results generated by virtual point sensors of different lengths so that our algorithm will automatically select suitable sampling methods for different types of path.The experiment shows that our algorithm performs well in scenes of difficult visibility containing mostly diffusive light incomings with a few glossy surfaces.