欢迎您！东篱公司
退出
-
申报数据库
-
立项数据库
-
成果数据库
-
项目获奖数据库
中文摘要：
提出了一种针对码书优化的图像矢量量化算法。首先设定矢量距离的初始门限值,基于空间划分对训练矢量聚类,找到矢量个数小于平均胞腔矢量数的胞腔;提取其聚类矢量后删除该胞腔,然后缩小距离门限值再次聚类提取,依次循环直到聚类矢量个数达到要求。将提取的聚类矢量作为初始码书,进行LBG算法码书设计,有效地改善了传统LBG算法依赖于初始码书而容易陷入局部最优的缺点,采用改进算法生成的码书更接近全局最优,加快了收敛速度。仿真结果证明了该方法的有效性。
英文摘要：
An optimized VQ codebook design algorithm of image compression is introduced in this paper.First of all,set the initial distance threshold,and cluster the train vectors base on space partition,collect the clustery vectors which has more train vectors in its cell lumens than average vector number,and delete these cell lumens.Secondly,lessen the distance threshold and cluster again to obtain more clustery vectors until the number of clustery vectors meet requirement.Finally,consider these clustery vectors as the initial codebook for LBG algorithm. Though the efficiency of traditional LBG algorithm depends on the selecting of initial codebook,we can still obtain a codebook which close to global optimal and have a faster rate of convergence after we improved the initial codebook.The validity is proved by the simulation result.
同期刊论文项目
同项目期刊论文
位置：