Research On Image Encoding Algorithm Of Vector Quantization

Along with the popularization of the computer and digital communication, digital signal analysis and processing technology is taken seriously more and more by people and get a rapid development. Now this technology is widely used in radar, communication, aerospace and industrial automation etc. Digital signal has two outstanding advantages. First of all, transmission and storage system of digital has strong anti-interference ability, good secrecy and high reliability advantages. Secondly, digital signal is easy to delete the redundant information. But because the data volume of media information such as images, video and sound is usually very large, so it takes great effect on memory storage capacity, communication channel bandwidth and computer processing speed. Therefore, it is necessary to encode the digital signal in digital communication.Vector quantization (VQ) is one of the most important methods of data compression. VQ is an effective lossy compression method. In the coding process, VQ uses the most matching codeword in the codebook instead of an input vector. The salient characteristics of VQ are big compression ratio, simple decoding process and maintain the image details well. Many academic fields are involved by vector quantization techniques. The research on VQ will bring fresh blood for these fields. For this reason, the research on VQ has the great significance no matter from the theoretical perspective or from the application perspective.The study in this paper is the encoding algorithm on image vector quantization. It includes two aspects: 1. Code book design and optimization. It mainly devotes to the optimization of iterative algorithm to improve the code book performance. 2. The optimization of code words search algorithm. It mainly devotes to find more effectively code word exclusion criterion to exclude impossible code words and accelerate the encoding process. All in all, the key work in this paper are as follows:First of all, the vector quantization technology development situation at home and abroad was analyzed in the beginning of this paper. And then the design methods of current main vector quantizers are introduced. Several of high-performance codebook design method and fast encoding algorithms based on inequality and transform domain are also introduced in this paper.Secondly, an improved LBG algorithm is proposed in the third chapter. It is an effective codebook design algorithm base on space division. By introducing a distance adjustment factorλ, the clustering vector will be decreased gradually with the increase of iteration times. Thus, the input vector set is refined and the initial codewords are spread well among the input vector set.Last, a fast encoding algorithm for vector quantization based on weighted Variance Inequality and Hadamard Transform (HT) is proposed in the forth chapter. A new inequality which combined Variance Inequality and HT is used. By this inequality, more non-similar codewords will be rejected and the search range will decrescent. The experiment results show that the efficiency of the encoding process is improved observably.