Texture Compression and Relighting of 3-D Color Objects Using Singular Value Decomposition
I-CHENG CHANG1, HSIN-YO CHOU2 AND CHUNG-LIN HUANG2 1Department of Computer Science and Information Engineering National Dong Hwa University Hualien, 974 Taiwan 2Department of Electrical Engineering National Tsing Hua University Hsinchu, 300 Taiwan
Synthesis of 3-D objects is essential to animation, computer games and virtual reality. This work presents a novel view-dependent mapping method by taking the advantages of image-based and model-based rendering techniques. The proposed approach utilizes singular value decomposition to compute the coefficient matrices of the data captured from different viewing directions and different illuminations. These coefficient matrices are then compressed by the three-dimensional discrete wavelet transform. The remaining data are further coded via quantization and entropy coding to minimize the size for storage. We can generate the appearance of an object from a new viewing direction under a new light source. Experimental results evaluate the performance by testing different materials and demonstrate the effectiveness of the proposed approach.