The Constant Radiosity Assumption Syndrome

Most current radiosity solutions to the diffuse global illumination problem suffer from inherent inaccuracies because of the built-in assumption that radiosity is constant across the area of each patch. This paper presents an improved progressive refinement radiosity algorithm that makes the above assumption unnecessary. Progressive radiosity repeatedly chooses a light source and determines its contribution to the radiosities on a finite set of locations in the environment. An improved method for numerically approximating these contributions is presented. The method takes into account the non-uniform radiosity distribution across the light source to yield accurate results. In addition to the enhanced accuracy of the resulting solutions, dropping the constant radiosity assumption allows using larger patches, reducing the number of iterations required to obtain plausible results.