基于Hopfield神经网络的图像矢量量化  

A Method of Image Vector Quantization Based on Hopfield Neural Network

作　　者：荣秋生[1] 潘梅森[1] 颜君彪[1] 

机构地区：[1]湖南文理学院计算机科学与技术系,湖南常德415000

出　　处：《微计算机信息》2007年第02X期301-302,204,共3页Control & Automation

基　　金：湖南省教育厅一般项目(06C582);湖南省教育厅重点项目(04A037)

摘　　要：矢量量化是图像压缩的重要方法。论文提出了基于Hopfield神经网络的图像矢量量化方法,该方法首先构造聚类表格;然后聚类表格按离散Hopfield神经网络串行方式运行;最后根据得到的最终码字集,对图像进行矢量量化。论文最后给出模拟实验和结果比较,结果表明该方法是有效的,生成的码本质量优于传统的LBG算法。Image vector quantization is an important method in the image compression field. This paper proposed a method of image vector quantization based on hopfield Neural Network.This method first built a clustering form; then the clustering form worked according to the asynchronous mode of discrete hopfield neural network; Finally carried on the image vector quanfization according to the final codebook set. The paper finally produced the experiments and the result comparison, the result indicated this method is effective and the codebook quality surpassed traditional the LBG algorithm.

关 键 词：矢量量化 码本 LBG算 HOPFIELD神经网络 

分 类 号：TP391[自动化与计算机技术—计算机应用技术] TN919[自动化与计算机技术—计算机科学与技术]