Alexander Moschig

Apelles: a Library for Gpu-based Real-time Global Illumination Alexander Moschig

The evolution of the programmable graphics pipeline has given birth to numerous algorithms which utilise GPU-enabled rasterisation techniques to approximate individual effects of global illumination in real-time. This thesis focuses on a library named Apelles in order to discover which of these algorithms can be combined to generate convincing global illumination effects in real-time. As a consequence of the huge variety of the available algorithms, this thesis concentrates on the generation of both geometric hard and soft shadows cast by opaque objects only. An arbitrary number of light sources can be set-up in a scene naturally. The supported types of light sources include directional lights, point lights, spot lights and area lights. According to area lights, two algorithms are supplied for generating soft shadows. Visually pleasing soft shadows exhibiting contact hardening can be generated in real-time. Physically-correct soft shadows can be generated if a considerably higher cost is acceptable. Additionally, a GPU-enabled parallel algorithm is supplied to optimise the set-up of light sources with respect to those objects within the view of the light source that are relevant for resolving occlusion efficiently. The concise and open design of Apelles inherently supports to include missing global illumination effects by combining library algorithms with existing algorithms.