Research On Precomputed Interactive Global Illumination

Global illumination provides viewers with important cues that enhance understanding of shapes, materials and spatial relationships in a virtual scene and can significantly improve the realism of computer synthesized images. While global illumination is desirable in many interactive applications, the computational expense has been limiting its use to off-line applications. Recently, many researches focus on integrating global illumination into interactive applications such as illumination design, virtual reality systems and computer games, and precomputation-based methods have been commonly used to reduce the runtime computation cost.In this dissertation, we will focus on developing precomputation-based rendering techniques to enable interactive global illumination. Two novel methods are introduced.In the first method, we sample the intersections of lines in a discretized line space with scene objects using a two-step scan conversion algorithm. Results are organized into a new visibility representation, namely intersection field, or i-field in short. At runtime, the i-field can be queried to accelerate the line-scene intersection computation that required in light energy propagation. By reformulating the rendering equation and decomposing illumination into indirect irradiance and a product of direction irradiance and visibility ration, we are able to develop a GPU-based implementation to further improve runtime performance. By leveraging the line space coherence of intesection samples, we also develop a run length encoding technique for i-field compression.In the second method, we represent low-frequency lighting, reflectance and visibility by spherical harmonics(SH) vectors. Scene geometry is modelled as a set of spheres; relatively few spheres capture the low-frequency blocking effect of complicated geometry. At runtime, we compute the product of visibility vectors for these blocker spheres as seen from each receiver point. Instead of computing an expensive SH product per blocker as in previous work, we perform inexpen-...