灰度共生矩阵纹理特征提取的Matlab实现

2012－03－202012－06－24“”FIB070335－B8－041981－。Matlab225300、、。GLCMMatlab。、GLCM。MatlabTP391．9A1673－629X201211－0169－03ImplementationofGrayLevelCo－occurrenceMatrixTextureFeatureExtractionUsingMatlabJIAOPeng－pengGUOYi－zhengLIULi－juanWEIXingTaizhouCollegeofNanjingNormalUniversityTaizhou225300ChinaAbstractThetexturefeatureextractionofimageisafoundationalworkforimagerecognitionandclassificationcontent－basedimagere-trievalimagedataminingetc．Thetexturefeatureofimageissignificantfordescriptionofimagecontent．Anditisahottopicinthere-searchofimagenow．Themethodoftexturefeatureextractionbasedongraylevelco－occurrencematrixGLCMwasstudied．Analyzedtheeffectofeachparameteronconstructingtheco－occurrencematrixandimplementedthefeatureextractionusingMatlab．TheanalyticalresultsprovidevaluablereferenceforcreatingGLCMbetterandextractingtexturefeaturesofspecifickindsofimages．Keywordsgraylevelco－occurrencematrixtexturefeatureMatlab01。1。。2070、、。、、HaralickGLCM2。Matlab。1Haralick1NLθ00、450、900、1350d234Haralick14。。。。3L=32d=52211201211COMPUTERTECHNOLOGYANDDEVELOPMENTVol．22No．11Nov．201214Brodatz。。5GLCM、6GLCM144、、、。。GLCM7。GLCLL、GLCHS、GLCIA。2、89。θdijPijdθ。fxyxyLLr、Lc、。fxy=ifx＇y＇=jPijd00=#xyx＇y＇∈LrLc×LrLc|x＇－x=0|y＇－y|=dfxy=ifx＇y＇=j1Pijd450=#xyx＇y＇∈LrLc×LrLc|x＇－x=dy＇－y=dorx＇－x=－dy＇－y=－dfxy=ifx＇y＇=j2Pijd900=#xyx＇y＇∈LrLc×LrLc||x＇－x|=dy＇－y=0fxy=ifx＇y＇=j3Pijd1350=#xyx＇y＇∈LrLc×LrLc|x＇－x=dy＇－y=－dorx＇－x=－dy＇－y=dfxy=ifx＇y＇=j4#。P＇ijdθ=Pijdθ/RR10。Haralick14111C1。C1=∑L－1i=0∑L－1j=0Pijdθ252C2。C2=∑L－1i=0∑L－1j=0i－j2Pijdθ63C3。C3=∑L－1i=0∑L－1j=0Pijdθ1+i－j274C4。C4=∑L－1i=0∑L－1j=0PijdθlgPijdθ83Mat-lab4×440～31。d=10°4°90°135°1Pij100=4210240010610012·071·22Pij1450=213012103102002014*40011001102222233Pij1900=6020042022220020Pij11350=4100122002410010Matlab7．0N=4×4L=4d=1θ=45°qC1ticpix=0011001102222233mn=sizepixL=4q=doublezerosLLfori=1m－1forj=1n－1qpixij+1pixi+1j+1+1=qpixij+1pixi+1j+1+1+1endendfori=2mforj=2nqpixij+1pixi－1j－1+1=qpixij+1pixi－1j－1+1+1endendR=2*L－1*L－1c1=0fori=1Lforj=1Lqij=qij/Rc1=c1+qij^2endendqc1tocqMatlab110－L－1。C1=0．11730．015。Matlab。q=0．11110．05560．166700．05560．11110．055600．16670．055600．1111000．11110c1=0．1173Elapsedtimeis0．015000seconds．4GLCMNLθd。1N。ROIRegionOfInterestGLCM。2LGLCM。3d。。4θ00、450、900、1350θθ。Matlab。、、。1．D．2007．2HaralickRMShanmngamKDinsteinI．Texturefeaturefor175·171·11Matlab2One－against－oneOne－against－restSegmentγ=0．0526C=45C=312C=4660．9081．8901．2340．6250．3770．32897．473%95．56%97．033%satimageγ=0．0278C=22C=45C=3904．6878．8908．7185．3432．782．70389．15%88．8%88．85%Vehicleγ=0．0278C=87C=45C=870．7180．760．8120．1710．1360．12591．096%89．8%91．781%4SVMSVM。。1VapnikV．ThenatureofstatisticallearningtheoryM．NewYorkSpringer－Verlag1995．2WestonJWatkinsC．Multi－classsupportvectormachinesC//Proc．ofESANN99．BrusselsBelgiums．n．1999．3KrebelU．PairwiseClassificationandSupportVectorMachinesM//AdvancesinKernelMethodsSupportVectorLearn-ing．CambridgeMAMITPress1999255－268．4BottouLCortesCDenkerJetal．ComparisonofclassifiermethodsacasestudyinhandwritingdigitrecognitionC//Proc．ofInt．Conf．onPatternRecognition．s．l．s．n．199477－87．5PlattJCristianiniNShawe－TaylorJ．LargeMarginDAG＇sforMulticlassClassificationM//AdvancesinNeuralInforma-tionProcessingSystems12．CambridgeMAMITPress2000547－553．6．J．200541724－26．7．J．2006634－37．8MadzarovGGjorgjevikjD．Multi－classclassificationusingsupportvectormachinesindecisiontreearchitectureC//ProceedingsofEUROCON．s．l．s．n．2009288－295．9LZhigangLDerenQQianqing．Ananalyticaloverviewofmethodsformulti－categorysupportvectormachinesJ．Jour-nalofcomputerengineeringandapplication200440710－13．10．SVMJ．2006232－35．11．J．20094517167－169．12．J．2005207746－749．13．SVMJ．200825113268－3270．檪檪檪檪檪檪檪檪檪檪檪檪檪檪檪檪檪檪檪檪檪檪檪檪檪檪檪檪檪檪檪檪檪檪檪檪檪檪檪檪檪檪檪檪檪檪171imageclassificationJ．IEEETransactionsonSystemsManandCybernetics197336768－780．3．J．2010274224－227．4．J．20092941018－1021．5EquitzWNiblackW．RetrievingImagesfromaDatabaseU-singTexture－algorithmsfromtheQBICSystemR．s．l．s．n．1994．6．D．2010．7TakahashiNIwasakiMKuniedaTetal．Imageretrievalu-singspatialintensityfeaturesJ．SignalProcessingImageCommunication2000161－245－57．8．J．2010466217－220．9．SVMJ．201020317－19．10．D．2009．11YaoHYLiBC．Generalizedco－occurrencematrixmethodforcontent－basedimageretrievalJ．ComputerEngineeringApplications200434698－101．·571·11