Efficient Hierarchical Refinement and Clustering for Radiosity in Complex Environments

Abstract

Generating accurate radiosity solutions of very complex environments is a time-consuming problem. We present a rapid hierarchical algorithm that enables such solutions to be computed quickly and efficiently. Firstly, a new technique for bounding the error in the transfer of radiosity between surfaces is discussed, incorporating bounds on form factors, visibility, irradiance, and reflectance over textured surfaces. This technique is then applied to the problem of bounding radiosity transfer between clusters of surfaces, leading to a fast, practical clustering algorithm that builds on the previous work of Sillion1. Volumes are used to represent clusters of small surfaces, but unlike previous algorithms, the orientations of surfaces inside each cluster are accounted for in both the error bound and radiosity transfer. This enables an accurate solution to be generated very efficiently, and results are presented demonstrating the performance of the algorithm on a variety of complex models, one containing almost a quarter of a million initial surfaces.