The Study Of Wavelet Image Classified Vector Quantization And Trellis Coded Quantization

Based on the analysis of image wavelet transformation and the space/frequency distributing characteristics of different subbands' coefficients, this dissertation fully exploits the following theories and methods: scalar quantization, vector quantization, trellis coded quantization, trellis coded vector quantization, vector classification, codebook expansion and weighted mean square error rule basing mankind visual characteristics,etc. From different angles of information amalgamation, It develops several innovative algorithms of image compression and coding, gives their realization schemes, and makes plentiful simulation tests. The results show their efficiency and validity.Chapter 2 first reviews the basic theory concerned with image wavelet transformation, which includes the wavelet multi-resolution analysis theory, the discrete wavelet transformation and the two dimension discrete wavelet transformation (Mallat algorithm), and analyzes the space and frequency distributing characteristics of image wavelet coefficients. Then it proposes a new fast zerotree encoded algorithm combing wavelet image zerotree data-structure and optimum quantization, designs a realization scheme and gives simulation results and their analysis.Chapter 3 particularly introduces vector quantization and its applications to wavelet image quantization. It gives three effective coding schemes, firstly it reviews the fundamental theory of vector quantization and its current algorithms, then it analyzes and summarizes characteristics of wavelet image quantization with VQ. Three commixed still image coding algorithms are proposed based on them and such ideas as: zerotree coding, WMSE (which is based on mankind visual characteristics), classified vector quantization with different vectorstructures and classification methods. The chapter gives the principium analysis, realization schemes and simulation results of the proposed algorithms. At last, conclusions are drawn on these algorithms.Based on scalar quantization of chapter 2,Chapter 4 introduces trellis coded quantization to improve quantization gain. Since the fundamental idea of TCQ originates from trellis coded modulation, in order to comprehend TCQ, it firstly introduces TCM. Also, because Viterbi decoding algorithm is the key of coding gain of TCM and TCQ, it particularly introduces convolutional coding and Viterbi decoding algorithm. Based on them it emphatically describes TCQ principium and analyzes its performance. Following on the heels of these it proposes wavelet image TCQ algorithm, and analyzes its principium and its TCQ algorithm realization. Finally, it gives simulation results and its analysis.Chapter 5 is the direct extension of chapter 2, namely it maps SQ field to VQ field. This chapter first introduces the principium of trellis coded vector quantization, then focuses on analyzing the algorithm of wavelet image classified weighted TCVQ and its realization process, also gives TCVQ application to wavelet image coding and its simulation results. Finally, it draws conclusions to the algorithms of this chapter.Using TCVQ and ideas of small codebook expansion and block coding, Chapter 6 proposes a new quantization precept making quantization on two dimension codebook space, that is 2DJTCVQ. It designs an algorithm applying 2DJTCVQ to image quantization, also gives analysis and simulation results of 2D-TCVQ in image space domain.