Perceptual refinement for hierarchical radiosity

Abstract

This dissertation explores the use of a simple model of the human visual system to yield a performance improvement with hierarchical radiosity. Hierarchical radiosity is a physically based rendering algorithm and hence makes no attempt to optimize computation for human perception. We used a model of the edge enhancement properties of the human visual system to produce a perceptually based refinement oracle for the hierarchical radiosity algorithm. Tests of the perceptual oracle shows that it allows the hierarchical radiosity algorithm to produce the same visual quality output in half the time and using half the memory compared to the same algorithm using the standard refinement oracle.