Markov-Chain Monte Carlo Sampling of Visibility Boundaries for Differentiable Rendering

Physics-based differentiable rendering requires estimating boundary path integrals emerging from the shift of discontinuities (e.g., visibility boundaries). Previously, although the mathematical formulation of boundary path integrals has been established, efficient and robust estimation of these integrals has remained challenging. Specifically, state-of-the-art boundary sampling methods all rely on primary-sample-space guiding precomputed using sophisticated data structures—whose performance tends to degrade for finely tessellated geometries. In this paper, we address this problem by introducing a new MarkovChain-Monte-Carlo (MCMC) method. At the core of our technique is a local perturbation step capable of efficiently exploring highly fragmented primary sample spaces via specifically designed jumping rules. We compare the performance of our technique with several state-of-the-art baselines using synthetic differentiable-rendering and inverse-rendering experiments.