基于顏色和紋理的圖像檢索.pdf

1、而普它手斜校先警碩士學(xué)位論文題c十x女，目些王趣笪塑鯉堡盟圈嬋撿盎一一一ImageRetrievalBa：sedonColorandTexture一一一一作者姓名黑基通一指導(dǎo)教師姓名、職務(wù)一王煎醴l救援一學(xué)科門類_工堂。學(xué)科、專業(yè)童避信皇至型器撞趔一提交論文13期一一一三QQ七生=且一AbstractContentBasedImageRetrieval(CBIR)isapopulartechniqueinthefieldofimagep

2、rocessingCBIRextractsvisualfeaturesastheretrievalfeatures，suchascolortexture，shapeetcGenerallyfeaturesoftheexampleimageneededareextractedfirstlyandthencomparedwiththefeaturesofotherimagesinthedatabaseAftertheresultisshow

3、edtotheuserBasedonresearchinrelativefieldincludingdetailedanalysisofCBlRarchitecture，featuresextractionmethodsandimagematchtechnique，twomethodsareproposedinthispaperWecombinedtwotechniquesofbasedonchromaticitymomentstech

4、niqueandstatisticaldistributionofwavelettransformationcoefficientsforcontent—basedimageretrieval(CBIR)First，onlyasmallnumberoffeatures，calledchromaticitymoments，arerequiredtocapturethespectralcontent(chrominance)ofallima

5、geOnlyasmallfixednumberofcompactfeaturesneedtobestoredtoeffectivelycharacterizethecolorcontentofanimageSecondlythewaveletcoefficientsprovideimportantdirectionalinformation，thestatisticsofimagewaveletcoefficientsisgeneral

6、izedGaussiandistributionThisapproachformsthefeaturevectoroftheimagebytakingadvantageofthestatisticaldistributionofwavelettransformationcoefficientsExperimentalresultshowthatanefficientimageretrievalmethodbasedonchromatic

7、itymomentstechniqueandstatisticaldistributionofwavelettransformationcoefficientsTheapproachesusetheLocalBinaryPattern(LBP、texturefeatureasthesourceofimagedescriptionThispaperpresentsanimprovedLBPhistogramspatiallyenhance

8、dLBP(SEL)histogramwithspatialinformationbydividingtheLBPlabeledimagesintokxkrc：gionsWecomparethemethodofSELandthemethodofdynamicregionsmatching。Asshownintheexperimentalresults，themethodofSELisbetterthanthewholeofLBPmetho