Exploiting the GPU power for image-based relighting and neural network.

Wei Dan.Thesis submitted in: October 2005.Thesis (M.Phil.)--Chinese University of Hong Kong, 2006.Includes bibliographical references (leaves 93-101).Abstracts in English and Chinese.Chapter 1 --- Introduction --- p.1Chapter 1.1 --- Background --- p.1Chapter 1.2 --- Our applications --- p.1Chapter 1.3 --- Structure of the thesis --- p.2Chapter 2 --- The Programmable Graphics Hardware --- p.4Chapter 2.1 --- Introduction --- p.4Chapter 2.2 --- The evolution of programmable graphics hardware --- p.4Chapter 2.3 --- Benefit of GPU --- p.6Chapter 2.4 --- Architecture of programmable graphics hardware --- p.9Chapter 2.4.1 --- The graphics hardware pipeline --- p.9Chapter 2.4.2 --- Programmable graphics hardware --- p.10Chapter 2.5 --- Data Mapping in GPU --- p.12Chapter 2.6 --- Some limitations of current GPU --- p.13Chapter 2.7 --- Application and Related Work --- p.16Chapter 3 --- Image-based Relighting on GPU --- p.18Chapter 3.1 --- Introduction --- p.18Chapter 3.2 --- Image based relighting --- p.20Chapter 3.2.1 --- The plenoptic illumination function --- p.20Chapter 3.2.2 --- Sampling and Relighting --- p.21Chapter 3.3 --- Linear Approximation Function --- p.22Chapter 3.3.1 --- Spherical harmonics basis function --- p.22Chapter 3.3.2 --- Radial basis function --- p.23Chapter 3.4 --- Data Representation --- p.23Chapter 3.5 --- Relighting on GPU --- p.24Chapter 3.5.1 --- Directional light source relighting --- p.27Chapter 3.5.2 --- Point light source relighting --- p.28Chapter 3.6 --- Experiment --- p.32Chapter 3.6.1 --- Visual Evaluation --- p.32Chapter 3.6.2 --- Statistic Evaluation --- p.33Chapter 3.7 --- Conclusion --- p.34Chapter 4 --- Texture Compression on GPU --- p.40Chapter 4.1 --- Introduction --- p.40Chapter 4.2 --- The Feature of Texture Compression --- p.41Chapter 4.3 --- Implementation --- p.42Chapter 4.3.1 --- Encoding --- p.43Chapter 4.3.2 --- Decoding --- p.46Chapter 4.4 --- The Texture Compression based Relighting on GPU --- p.46Chapter 4.5 --- An improvement of the existing compression techniques --- p.48Chapter 4.6 --- Experiment Evaluation --- p.50Chapter 4.7 --- Conclusion --- p.51Chapter 5 --- Environment Relighting on GPU --- p.55Chapter 5.1 --- Overview --- p.55Chapter 5.2 --- Related Work --- p.56Chapter 5.3 --- Linear Approximation Algorithm --- p.58Chapter 5.3.1 --- Basic Architecture --- p.58Chapter 5.3.2 --- Relighting on SH --- p.60Chapter 5.3.3 --- Relighting on RBF --- p.61Chapter 5.3.4 --- Sampling the Environment --- p.63Chapter 5.4 --- Implementation on GPU --- p.64Chapter 5.5 --- Evaluation --- p.66Chapter 5.5.1 --- Visual evaluation --- p.66Chapter 5.5.2 --- Statistic evaluation --- p.67Chapter 5.6 --- Conclusion --- p.69Chapter 6 --- Neocognitron on GPU --- p.70Chapter 6.1 --- Overview --- p.70Chapter 6.2 --- Neocognitron --- p.72Chapter 6.3 --- Neocognitron on GPU --- p.75Chapter 6.3.1 --- Data Mapping and Connection Texture --- p.76Chapter 6.3.2 --- Convolution and Offset Computation --- p.77Chapter 6.3.3 --- Recognition Pipeline --- p.80Chapter 6.4 --- Experiments and Results --- p.81Chapter 6.4.1 --- Performance Evaluation --- p.81Chapter 6.4.2 --- Feature Visualization of Intermediate-Layer --- p.84Chapter 6.4.3 --- A Real-Time Tracking Test --- p.84Chapter 6.5 --- Conclusion --- p.87Chapter 7 --- Conclusion --- p.90Bibliography --- p.9