Globally illuminated anisotropic lighting models with an enhanced hemi-cube sampling techinque on a GPU.

摘要

Global illumination is a very important component of rendering photorealistic images. It simulates light inter-reflection by sampling the light paths between object surfaces. This process consists of two parts --- direct illumination and indirect illumination. The majority of the rendering time is incurred in calculating indirect illumination, which is a computationally expensive step. Since global illumination can produce many realistic effects such as glossy reflections and caustics, many researchers continue to develop hybrid global illumination approaches to reduce the time required in computing indirect illumination. With the rapid development of graphics processing units (GPU), many global illumination techniques are adapted or modified so that they can be implemented on the GPU to achieve lower rendering times. Among these GPU-based global illumination methods, the lighting model for anisotropic materials has not yet been covered extensively. The goal of this thesis is to implement global illuminated anisotropic materials on the GPU. The main algorithm of this thesis employs the irradiance cache algorithm to compute the indirect illumination. The hemi-cube technique is used in the sampling step of the algorithm. Due to the fact that the hemicube only approximates the hemisphere above a hit point, a sharp edge cut-off artifact occurs and becomes severe for anisotropic materials. This thesis introduces a novel approach --- multiple fields of view --- to solve this problem. The model built here balances the workload between the CPU and the GPU and utilizes the computational power of the GPU. It can generate high quality global illuminated scenes with anisotropic materials in a reasonable amount of time. It also broadens the usage of the irradiance cache splatting algorithm to a wider range of applications.