The Research Of Vector Quantization Based On Wavelet Transform On Image Coding

Vision is important in human's sensation. Image, as the result of human vision sensation, is the most important information source for human to realize the world. So, image has now become the most important data type in the filed of multimedia technology, and image process and analysis technology has now become the specialty in the field of modern signal process. Because of the great amount of data and information in image, it is necessary to compress the image data in order to adapt to the practical application demand. The redundancy of statistic, structure and vision in image provides the possibility to image compress. Image Code is a developing field on image compress and a lot of methods on that has now generated, such as statistic code, predict code, transform code, subband code, model code, wavelet transform code, vector quantization code, neural networks code, fractal code etc.In this thesis, we mainly describe our work on vector quantization code and wavelet transform code on image compress. After summarize the basic theory on image code compress, we firstly introduce the vector quantization theory, and we also propose a LBG improved algorithm based on simulated annealing with experimentation to verify the performance. Secondly, we introduce the wavelet transform theory and the EZW algorithm used in the area of image compression. Based on the former two theories, we propose a wavelet transform cross-band vector quantization based on human vision, and we verify the correction with experimentation. At last, in view of parallel, we design a 2-dimension DCT image compress code algorithm based on data partition, and verify the algorithm's performance in the PVM environment.