Importance-Driven Progressive Refinement Radiosity

A simple way of using importance, also called visual potential, in progressive refinement radiosity and similar radiosity methods is presented. This is accomplished by taking the visual potential of the patches into account when choosing of the “next most contributing” patch. Quite faster convergence is obtained for individual views in which only a relatively small part of a scene is visible. We also show how changes of the viewing parameters can be handled efficiently by computing importance incrementally.