图像块动态划分矢量量化  

A Novel Method of Vector Quantization with Automatic Image Partition

作　　者：马文龙[1] 余宁梅[1] 银磊[1] 高勇[1] 

机构地区：[1]西安理工大学电子工程系,西安710048

出　　处：《计算机辅助设计与图形学学报》2005年第2期279-283,共5页Journal of Computer-Aided Design & Computer Graphics

基　　金：国家自然科学基金 (60 2 760 17)

摘　　要：提出一种可以自动调节子图像块尺寸的矢量量化算法———图像块动态划分矢量量化 该算法在进行编码前 ,首先对待编码子图像块及其相邻块进行相关度分析 ,通过预设的阈值 (最大相关度 )进行分类编码 :如果相关度小于阈值 ,则对 8× 8图像块进行编码 ;否则 ,对 4× 4的图像块进行编码 实验结果表明 ,相对于普通矢量量化 ,文中算法不但可以提高压缩率和编码速度 。This paper describes a novel vector quantization (VQ) algorithm called partition dynamically VQ (PDVQ),capable of adjusting the size of image block automatically. By the method, before encoding,PDVQ first analyzes the correlation of the encoding image block and its neighboring blocks, then decides the encode method according to the comparison of the correlation result with the threshold value(the maximum correlation,MC) given by experiment. If the correlation result is less than MC, an image block size of 8×8 will be encoded, otherwise, an image block size of 4×4 will be encoded. Test result shows that PDVQ can improve the compression rate and coding speed in contrast with the normal VQ, and more importantly, it will not effect the quality of the image reconstructed.

关 键 词：矢量量化 动态划分 相关度 最大相关度 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]