Radiosity is an advanced technique in generating photo-realistic images which is useful in applications involving illumination engineering, architectural design, interior decoration, as well as the study of radiative properties of materials. The first step in using a radiosity renderer is to provide a discretized environment as the input, which is usually created by some mesh generators. In the first part of this thesis, the advancing front mesh generation algorithms are studied. These are popular in fluid mechanics and are able to triangulate arbitrary polygonal surfaces by maintaining the Delaunay Triangulation properties. Improvement based on grid searching is proposed to accelerate the mesh generation process. In the second part of the thesis, five radiosity algorithms (PR, Shao, Go...[ Read more ]

Radiosity is an advanced technique in generating photo-realistic images which is useful in applications involving illumination engineering, architectural design, interior decoration, as well as the study of radiative properties of materials. The first step in using a radiosity renderer is to provide a discretized environment as the input, which is usually created by some mesh generators. In the first part of this thesis, the advancing front mesh generation algorithms are studied. These are popular in fluid mechanics and are able to triangulate arbitrary polygonal surfaces by maintaining the Delaunay Triangulation properties. Improvement based on grid searching is proposed to accelerate the mesh generation process. In the second part of the thesis, five radiosity algorithms (PR, Shao, Gortler, Feda, and Xu) are investigated under a mathematical framework. These algorithms can all be integrated under a general linear iteration solver by using different parameters. By analysis of these parameters, performance prediction can be made and verified experimentally. The final part of the thesis is on the calculation of the form factor which is probably the most time-consuming and error-prone step in radiosity rendering. Problems with the convergence of algorithms due to the errors in form factor computations are identified and a solution is proposed.