Efficient Radiosity for Complex Environments
Simon Gibson.
M.Sc. Thesis, University of Manchester, September 1995.
Abstract
The radiosity method has become one of the major tools used by computer graphics researchers to generate realistic images, due to its ability to simulate the interreflection of light between diffuse surfaces. For geometrically complex environments, the enormous number of surface-to-surface interactions that must be computed to generate an acurate solution soon becomes overwhelming, thereby limiting the use of radiosity in practical applications.
This thesis focuses on recently developed hierarchical radiosity techniques that are capable of reducing both memory requirements and computation time. The basic hierarchical algorithm is discussed, and its shortcomings identified. This leads to the development of a more efficient error-driven refinement strategy, and results are given showing the superiority of this technique over the standard method. Also, by combining this refinement strategy with a hierarchical texture representation, it is shown how emissive and reflective textures may be accurately incorporated into the radiosity simulation. Following this, two methods for accelerating the overall solution process are compared - lazy link evaluation and clustering. An improved error-driven clustering algorithm is presented, and results given showing how very complex radiosity solutions can be computed quickly and efficiently.