基于树结构矢量分类的小波图像网络编码矢量量化  

Trellis coded vector quantiztion of wavelet image using tree structure vector combination and classification

作　　者：郑勇[1] 周正华[1] 朱维乐 

机构地区：[1]电子科技大学电子技术系,四川成都610054

出　　处：《通信学报》2001年第9期108-114,共7页Journal on Communications

基　　金：中兴通讯研究基金资助项目

摘　　要：本文基于零树编码、矢量分类和网络编码量化的思想 ,提出了对小波图像采用树结构矢量组合和分类后进行网格编码矢量量化的新方法。该方法充分利用了带系数的带间和带内的相关性 ,分类信息占用比特数少 ,对重要类矢量实行加权网格编码矢量量化 ,利用卷积编码扩展信号空间以增大量化信号间的欧氏距离 ,用维特比算法搜索最优量化序列 ,并采用基于人眼视觉特性的加权均方误差准则作为失真度量和码字匹配 ,提高了量化增益。仿真结果表明 ,该方法编码计算复杂度适中 ,解码简单 。Based on zerotree coding,vector classification and trellis coded quantizaion,A new method which performs tree structure vector formation and classification to wavelet image then uses TCVQ is proposed the method utilizes correlation between the subbands and within the subbands fully,the classified information need a little bites to express In order to improve quantizaiton gain,the important vectors are quantified by weighted TCVQ,which uses convolutional coding to expand signal space and increase Euclid distance between signals,uses Viterbi algorithm to find a optimized survived quantized sequence and adopts weighted mean square error rule to that using mankind visual characteristic to measure the distortion Simulation states that this method has modest encoding complexity with simple decoding and it can achieve much better compression effect

关 键 词：小波变换 树结构矢量 网格编码矢量量化 图像编码 

分 类 号：TN919.81[电子电信—通信与信息系统]