Tomoya Yamaguchi, Tatsuya Yatagawa, Shigeo Morishima

Efficient Metropolis Path Sampling for Material Editing and Re-rendering

Pacific Graphics 2018, Short Paper

http://sweb.cityu.edu.hk/pg2018/

This paper proposes efficient path sampling for re-rendering scenes after material editing. The proposed sampling is based on Metropolis light transport (MLT) and dedicates path samples more to pixels receiving greater changes by editing. To roughly estimate the amount of the changes in pixel values, we first render the difference between images before and after editing. In this step, we render the difference image directly rather than taking the difference of the images by separately rendering them. Then, we sample more paths for the pixels with larger difference values, and render the scene after editing following a recent approach using the control variates. As a result, we can obtain fine rendering results with a small number of path samples. We examined the proposed sampling with a range of scenes and demonstrated that it achieves lower estimation errors and variances over the state-of-the-art method.