Lazy indirect illumination evaluation using barycentric coordinates

Abstract

Global illumination algorithms usually spend the majority of time on visibility queries by casting enormous numbers of shadow rays in a three‐dimensional scene containing area light sources. In this paper, we present a technique to accelerate visibility queries for indirect illumination in Monte Carlo integration. In contrast to the standard approach, our technique performs visibility queries merely on vertices of solid angles subtended by luminaires. Barycentric coordinates are then employed to approximate the visibility between a shaded point and an area light source. This avoids elaborate and expensive visibility examinations during stochastic procedures. Results show that visibility approximation produces a roughly similar image quality in significantly less time than standard visibility queries. In addition, our technique can greatly decrease the number of visibility queries even as the model complexity increases. The visibility approximation technique is simple, efficient, and feasible for quickly producing global illumination images of high quality.