作　　者： ; ; ;

机构地区： 华南理工大学电子与信息学院

出　　处： 《小型微型计算机系统》 2009年第4期780-783,共4页

摘　　要： 目前对传统LBG算法的改进措施一般以增加时间开销作为代价.本文提出一种新的矢量量化码书设计改进措施——初始码字间距最大化:初始码书中的码字全部来自输入的训练矢量,且每一个新的初始码字尽可能地远离现有的码字.实验结果表明:本算法完全消除了空胞腔现象,更有效地避免了局部最优,能获得质量更高的码书;收敛速度快,具有较低的时间消耗.本算法在时间开销以及码书质量这两个方面都优于传统LBG和基于人工蚁群优化的码书设计算法等改进算法. Many improvements on LBG algorithm are achieved at the expense of more runtime. This paper presents a novel improvement on codebook design for image vector quantization with the most dispersed codewords in initialization （MDCI）. In MDCI, all the initial codewords are selected from the inputted training vectors set, and the distance between the next newly generated initial codeword and the already existed codewords must be the greatest. Experimental results demonstrate MDCI conquers the empty cell problem, alleviates the problem of local optima more effectively and gets higher-performance codebook, outperforming the conventional LBG algorithm and many LBG-based modified algorithms, like the ant colony optimization based codebook design algorithm with respect to both codebook performance and runtime.

关 键 词： 图像压缩 矢量量化 码书 初始码字 间距最大化

领　　域： [电子电信] [电子电信]