Texture classification using spectral histograms
IEEETRANSACTIONSONIMAGEPROCESSING,VOL.12,NO.6,JUNE2003661
TextureClassificationUsingSpectralHistograms
XiuwenLiu,SeniorMember,IEEE,andDeLiangWang,SeniorMember,IEEE
Abstract—Basedonalocalspatial/frequencyrepresentation,weemployaspectralhistogramasafeaturestatisticfortextureclas-sification.Thespectralhistogramconsistsofmarginaldistribu-tionsofresponsesofabankoffiltersandencodesimplicitlythelocalstructureofimagesthroughthefilteringstageandtheglobalappearancethroughthehistogramstage.Thedistancebetweentwospectralhistogramsismeasured>>>>using2-statistic.Thespec-tralhistogramwiththeassociateddistancemeasureexhibitssev-eralpropertiesthatarenecessaryfortextureclassification.Afilterselectionalgorithmisproposedtomaximizeclassificationperfor-manceofagivendataset.Ourclassificationexperimentsusingnat-uraltextureimagesrevealthatthespectralhistogramrepresenta-tionprovidesarobustfeaturestatisticfortexturesandgeneralizeswell.Comparisonsshowthatourmethodproducesamarkedim-provementinclassificationperformance.Finallywepointouttherelationshipsbetweenexistingtexturefeaturesandthespectralhis-togram,suggestingthatthelattermayprovideaunifiedtexturefeature.
IndexTerms—Filtering,spectralhistogram,textureanalysis,textureclassification.
I.INTRODUCTION
EXTUREclassificationisafundamentalproblemincom-putervisionwithawidevarietyofapplications[40].Twofundamentalissuesintextureclassificationarehowtocharac-terizetexturesusingderivedfeaturesandandhowtodefinearobustdistance/similaritymeasure