Many-Worlds Inverse Rendering

TL;DR

This paper introduces a novel many-worlds inverse rendering approach that models multiple conflicting explanations of a scene using volumetric perturbations, resulting in a simpler and more efficient Monte Carlo algorithm for surface optimization.

Contribution

It proposes a new many-worlds representation differentiating volumetric perturbations, improving efficiency and convergence in physically-based inverse rendering.

Findings

Faster convergence in surface optimization tasks.

Simpler Monte Carlo algorithm compared to prior methods.

Effective handling of visibility changes in inverse rendering.

Abstract

Discontinuous visibility changes remain a major bottleneck when optimizing surfaces within a physically-based inverse renderer. Many previous works have proposed sophisticated algorithms and data structures to sample visibility silhouettes more efficiently. Our work presents another solution: instead of differentiating a tentative surface locally, we differentiate a volumetric perturbation of a surface. We refer this as a many-worlds representation because it models a non-interacting superposition of conflicting explanations (worlds) of the input dataset. Each world is optically isolated from others, leading to a new transport law that distinguishes our method from prior work based on exponential random media. The resulting Monte Carlo algorithm is simpler and more efficient than prior methods. We demonstrate that our method promotes rapid convergence, both in terms of the total iteration count and the cost per iteration.