关于二维主成分分析方法的研究

？Theprincipalcomponentanalysis(PCA),ortheeigenfacesmethod,isadefactostandardinhumanfacerecognition.NumerousalgorithmstriedtogeneralizePCAindifferentaspects.Morerecently,atechniquecalledtwo-dimensionalPCA(2DPCA)wasproposedtocutthecomputationalcostofthestandardPCA.UnlikePCAthattreatsimagesasvectors,2DPCAviewsanimageasamatrix.Withaproperlydefinedcriterion,2DPCAresultsinaneigenvalueproblemwhichhasamuchlowerdimensionalitythanthatofPCA.Inthispaper,weshowthat2DPCAisequivalenttoaspecialcaseofanexistingfeatureextractionmethod,i.e.,theblock-basedPCA.UsingtheFERETdatabase,extensiveexperimentalresultsdemonstratethatblock-basedPCAoutperformsPCAondatasetsthatconsistofrelativelysimpleimagesforrecognition,whilePCAismorerobustthan2DPCAinhardersituations.