Vector Quantization And Its Application In Image Compression,

Vector quantization(VQ) is an effective lossy compression technique with prominent merits of high compression rate and easy decoding algorithm. As the relevant theoretical and applied research has not been perfect, the VQ compression standard has not been build at present. This paper has deeply researched two key techniques of VQ,codebook design and codeword search, and image compression algorithm based on vector quantization.In codebook design algorithm respect,for defects of classic LBG algorithm heavily dependent on the initial codebook and the global optimization algorithm large computation,a new algorithm to generate vector quantization initial codebook is proposed, Component Mean Orthogonal Segmentation Algorithm(CMOSA). CMOSA orthogonally segments sample space using component mean to make full use of codeword. The solution for atypical codeword is proposed,so that the new algorithm is also suited for non-uniform distribution samples. The simulation results indicate that,between training speed and quantization distortion of the two conflicting indicators,CMOSA achieves a better balance than the previous algorithms.In codecode search algorithm respect, This paper focuses on codecode search algorithm based on inequality principle and mean variance principle. A modified equal-mean and equal-variance nearest neighbor search is proposed,which has combined the range advantage of absolute error inequality principle with the speed advantage of mean variance principle. The methods to selecting initial match codeword and to deciding searching order were proposed. The simulation results show that the searching speed has be improved by approximately 10% compared with the former.In image compression application respect, for corelation side match vector quantization can not adaptively adjust bit-rate,Two methods,mean method and variance method, have been proposed to identify "high-detail block" and "low-detail block". High-detail blocks which include rich detail information are coded with higher bit-rate,or with lower bit-rate. The simulation results show that this adaptive variable bit-rate VQ image compression algorithm has improved more in the three important performance indicators of codeing quality,bit-rate,and coding time. Combining huffman coding on side match vector quantization,a new algorithm is proposed. The simulation show that, bit-rate of this algorithm approachs to the lower limit of the information entropy theory,while the distortion does not increase. It may become a referent method to build international standard of image compression based on vector quantization.