首页 | 本学科首页   官方微博 | 高级检索  
     

基于PCA/SOFM混合神经网络的矢量量化
作者单位:辽宁工业大学电子与信息工程学院
摘    要:针对自组织特征映射(SOFM)神经网络应用于矢量量化具有收敛速度慢、计算量大等缺点,本文提出了一种基于PCA/SOFM混合神经网络的矢量量化的算法,先用主元分析(PCA)线性神经网络对输入矢量进行降维处理,再用SOFM神经网络进行矢量量化。通过调整SOFM神经网络的学习函数、邻域权值及初始码书对网络进行优化。实验表明,改进算法缩短了图像压缩的时间,提高了码书的性能。

关 键 词:矢量量化  自组织特征映射神经网络  图像压缩  主元分析

The Vector Quantization Based on PCA/SOFM Hybrid Neural Network
Authors:HUNG Cui-cui  ZHANG Jian
Abstract:In order to improve the two main shortcomings of the Kohonen's self-organizing feature map(SOFM) that are high computa-tion complexity and poor codebook quality,the author proposes a vector quantization algorithm based on PCA/SOFM hybrid neural net-work in this paper.Descend the dimension of imported vectors by using the principal component analysis(PCA) linear neural network.And then,use SOFM neural network to vector quantization.By modifying the learning-rate parameter,topology field weight and initial codebook of the SOFM neural network to optimize network.Simulation results demonstrate that the image compression algorithm can shorten the time and improve the performance of codebook.
Keywords:Vector quantization(VQ)  Self-organizing feature map neural network(SOFM)  image compression  Principle component analysis(PCA)
本文献已被 CNKI 等数据库收录!
设为首页 | 免责声明 | 关于勤云 | 加入收藏

Copyright©北京勤云科技发展有限公司  京ICP备09084417号