基于NMF和LPP的降维方法  

Dimensionality Reduction Method Based on NMF and LPP

作　　者：董焕[1] 闫德勤[1] 

机构地区：[1]辽宁师范大学数学学院,辽宁大连116029

出　　处：《吉林师范大学学报（自然科学版）》2011年第4期60-63,共4页Journal of Jilin Normal University:Natural Science Edition

基　　金：中国科学院自动化研究所复杂系统与智能科学重点实验室开放课题基金(20070101);辽宁省教育厅高等学校科学研究基金(2008344)

摘　　要：NMF是一种近年来常用的降维方法.NMF在图像检索、人脸识别和信号处理等方面得到广泛的应用,其分解后所产生的分量的非负性要求,使数据处理得到很好的效果.NMF在分解过程中未考虑到数据的内在几何性质和局部结构,就存在着不能准确的处理数据的问题.本文提出一种把NMF与LPP相结合的降维方法.该方法应用在图像检索上,因为LPP能够保留数据的内在几何性质和局部结构,降低影响图像检索的的因素,从而提高了图像检索的效率.再从Corel数据库进行实验,来证明此方法确实能够提高了检索准确性.NMF is a commonly used dimensionality reduction method in recent years which is also widely used in image retrieval,face recognition and signal processing.The weight＇s non-negative demand produced by decomposition made the data processing a good result.NMF in the process of decomposition without considering the geometric properties and local structure internal data,so there could be some problems in processing the data accurately.This paper puts an innovative approach which combines NMF and LPP together into image retrieval use in that LPP retains the local structure and geometric properties of data.Not only decreases negative effect in image retrieval,this method could also make signal processing more efficient.A verification based on corel database has proved that the method can indeed improve the retrieval accuracy.

关 键 词：NMF(非负矩阵分解) LPP(局部保留映射) 图像检索 K-MEANS聚类 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]