针对原有随机数设置法、训练矢量集随机抽取法和LBG分裂法等初始码书算法存在码矢利用率较低、运算量大和与信源匹配程度不高等不足,提出了一种新的分离平均法,并应用到基于自组织特征映射算法(SOM)的学习矢量量化(LVQ)中.图像矢量量化的实验表明,分离平均初始码书算法具有无效码矢数量少和码书性能高、运算量小、实现简单等优点.
In vector quantization(VQ), the initial codebook design is very important for VQ codebook performances. To overcome disadvantages of existing initial codebook algorithms, a new separating mean algorithm for learning vector quantization(LVQ)based upon self-organizing feature maps(SOM) was proposed. Experimental results for image VQ show that new initial codebook algorithm is better than random and splitting algorithm.