To split or not to split: case studies on Monte Carlo analysis of illumination ray tracing concerning the usefulness of ray-splitting

Abstract

In Monte Carlo ray tracing, the efficacy of variance reduction techniques is often the subject of debate. One portion of the debate regards the use of ray-splitting in illumination analysis. While analysis results should be the same whether raysplitting or no ray-splitting is used, one approach might result in better precision for a given calculation time. Additionally, inexperienced illumination designers may perform analyses in such a way as to exacerbate the difference in precision between the two methods. This results in a very important decision for illumination designers: using ray-splitting or not can affect results and the time spent getting them. For this paper, common illumination applications are analyzed by ray tracing in TracePro1 (a non-sequential Monte Carlo ray tracing program) and the analysis results are compared. Both raysplitting and no ray-splitting methods are used to see if the analyses converge to the same results for simple setup conditions. The results will illustrate the factors to consider of before choosing to use ray-splitting or not, and show examples when one method may be better than another.