A parallel radiosity system for large data sets

The radiosity method gives realistic results for synthesising images based on diffuse light interaction; however, the method is also very computationally expensive. Here we present a system for parallelising a progressive refinement radiosity method using a ray casting approach. The system decomposes the environment into two distinct entities, the model geometry and the radiosity sample points. These are then distributed amongst the processors and message passing communication protocols and caching methods are introduced to enable the system to cope with very large models while maintaining a high processing rate