Multidimensional cubic lattice vector quantization for wavelet transform image coding

Abstract

This thesis proposes an image coding method for the visual telecommunication system used in the very low bitrate channel capacity with the less loss of image quality. In order to achieve the major purposes that are low entropy and low distortion, the system must pay expensiveness in terms of the computation time and memory. In this thesis, the uniform cubic lattice is chosen as codebook of Lattice Vector Quantization (LVQ)``s codebook, because of its generic simplicity. On the other hands, the transform of image is important to apply on the LVQ. In the field of transform coding, the Discrete Cosine Transform(DCT) is generally used because of its excellent energy compaction and fast algorithm, recently however, the Discrete Wavelet Transform(DWT) is actively researched for the reason of its multiresolution property. This proposed algorithm is basically composed of the biorthogonal DWT and the uniform cubic LVQ. In this proposed algorithm, the multiresolution property of the DWT is actively used to solve the problem of the entropy and the distortion on the base of the distortion-rate function and the entropy function. Thus, the uniform cubic LVQ is designed to have the mutiresolution codebooks. The codebooks are also designed to be optimal at each subimage that is analyzed by the biorthogonal DWT. For this reason, the codebook must have the different dimension depending on the subimage``s variance. The simulation results shows that the performance of the proposed coding method is superior to others in terms of the entropy, the distortion, the number of operations and storage cost in this paper.