Research On Realistic Rendering Method Based On Metropolis

Rendering of complex global illumination scenes has always been a difficult problem in graphics rendering.The Metropolis Light Transport algorithm can gradually converge from an arbitrary distribution to a steady-state distribution,not only can efficiently render complex scenes,but also has good unbiasedness and robustness.After finding an effective light path,the Metropolis Light Transport algorithm can quickly mutate near the light path to find other effective light paths of the same length.Selecting an arbitrary distribution initially can effectively reduce the time taken to select the distribution at the time of rendering,and a fixed-length light path also facilitates the calculation of the final radiometric value.Although these two features improve the efficiency of the algorithm,but also affect the quality of the rendering results.For this reason,two improved methods are proposed.One is an improved Multiplexed Metropolis Light Transport algorithm based on variance filtering.The algorithm mainly aims at improving the start-up bias problem when Multiplexed Metropolis Light Transport algorithm chooses any distribution to converge,and better solves the problem of high variance of Multiplexed Metropolis Light Transport algorithm from initial sampling to formal sampling.The other is the adaptive optical path selection Multiplexed Metropolis Light Transport algorithm.Improved the poor rendering effect of the Multiplexed Metropolis Light Transport algorithm due to the fixed optical path length when dealing with caustics and high light scenes.By changing the optical path length in a targeted manner,the Multiplexed Metropolis Light Transport algorithm can effectively improve the processing of these two scenarios rendering effect.Experiments show that the first improved algorithm can reach the steady state distribution faster than the Primary Sample Space MLT algorithm and MMLT algorithm.The second algorithm can better handle the smoothing surface object scene.Finally,the two algorithms are fused,variance filtering is performed in the pre-sampling stage,adaptive optical path selection is used in the formal sampling stage,and the fusion algorithm is compared with the previous algorithm.The results show that the fusion algorithm has better experimental results than the PSSMLT and MMLT algorithm.