Global Illumination based on Antiradiance

(01.01.2006 - 30.04.2010)

Finite element methods for global illumination require enormous computation times, the majority of which is used to compute mutual visibility between surface patches. By using the novel concept of "Antiradiance" this visibility determination is eliminated by a simple trick: occlusion is ignored, but instead surfaces emit negative light (Antiradiance) generating their own shadow. The solution of the global illumination problem is thus significantly simplified, however at the cost of increased memory consumption and additional discretization artifacts. Goal of the project is the examination of this approach, a deeper understanding of the introduced discretization problems, and an efficient implementation on modren parallel hardware.