摘要
自组织特征映射（SOFM）是一种常用的矢量量化算法，它具有设计码书不依赖于初始码书等优点。模糊矢量量化算法（FVQ）将模糊关系引入码书的设计，训练矢量与码矢之间的模糊关系用隶属函数表示。本文提出了一种基于自组织特征映射网络的模糊矢量量化算法（FSOFM），FSOFM算法将SOFM网络的调节节点邻域看作训练矢量的模糊集，网络权值学习步长的选择依赖于隶属函数。由于设计码书的评价一般采用最小均方误差准则，而隶属函数是训练矢量与码矢之间距离的函数，FSOFM算法保证了网络的全局成优化和网络权值的局部调整一致；因此，FSOFM算法能够优化码书的设计，改善设计码书的性能。此外，FSOOFM算法还具良好的适应性，当网络的将LBG、SOFM、FVQ和FOSOFM算法用于一组具有不同边缘特性的图像的矢量量化中，我们发现采用FSOFM算法进行矢量量化的所有图像都具有最高的峰值信噪比PSNR。
Fuzzy vector quantization algorithms (FVQ) allow that each theining vector is assigned to multiple clusters, and a training vector translates from the fuzzy mode to the crisp mode according to an assignment strategy in a codebook design proccess. FVQ reduces the dependence of the resulting codebook on the random initial codebook selection, but FVQis not an iterative algorithm, because a codebook is modified afer all training vectors complete once iteration. This paper presents iterative fuzzy vector quantization algorithm (FSOFM) based on Self - Organizing Feature Map net, which inherits the advantages of SOFM and proposes a new strategy of the net weights modificatio.Such a strategy modifies the net wieghts according to a propsoed membership function and a training vector which just complete an iteration. In this paper, LBG, FVQ, SOFM, FSOFM are used in image compression based on vector quantization. We evaluate the computational efficiency of the proposed algorithms and the quality of the resulting codebook, and conclude that FSOFM is much better than other algorithms in many fields.
出处
《微型电脑应用》
1997年第4期40-43,共4页
Microcomputer Applications
关键词
码书
码矢
模糊系统
矢量量化
算法
FVQ
codebook codevector fuzzy relation fuzzy set membership function SOFM
