资源描述：
《irm integrated region matching for image retrieval》由会员上传分享，免费在线阅读，更多相关内容在教育资源-天天文库。
1、IRM:IntegratedRegionMatchingforImageRetrievalJiaLiyJamesZ.WangzGioWiederholdPaloAltoResearchCenterDept.ofComputerScienceDept.ofComputerScienceXeroxCorporationStanfordUniversityStanfordUniversityPaloAlto,CA94304Stanford,CA94305Stanford,CA94305jiali@db.stanford.ed
2、uwangz@cs.stanford.edugio@cs.stanford.eduABSTRACTdistributionproperties.Thistrendhasmotivatedresearchinimagedatabases,whichwerenearlyignoredbytraditionalContent-basedimageretrievalusingregionsegmentationhascomputersystemsduetotheenormousamountofdatanec-beenanacti
3、veresearcharea.WepresentIRM(IntegratedessarytorepresentimagesandthedicultyofautomaticallyRegionMatching),anovelsimilaritymeasureforregion-analyzingimages.Currently,storageislessofanissuesincebasedimagesimilaritycomparison.Thetargetedimagehugestoragecapacityisava
4、ilableatlowcost.However,ef-retrievalsystemsrepresentanimagebyasetofregions,fectiveindexingandsearchingoflarge-scaleimagedatabasesroughlycorrespondingtoobjects,whicharecharacterizedbyremainsasachallengeforcomputersystems.Theauto-featuresre
ectingcolor,texture,shap
5、e,andlocationprop-maticderivationofsemanticsfromthecontentofanimageerties.TheIRMmeasureforevaluatingoverallsimilarityisthefocusofinterestformostresearchonimagedatabases.betweenimagesincorporatespropertiesofalltheregionsinImagesemanticshasseverallevels:semantictyp
6、es,objecttheimagesbyaregion-matchingscheme.Comparedwithcomposition,abstractsemantics,anddetailedsemantics.retrievalbasedonindividualregions,theoverallsimilarityapproachreducesthein
uenceofinaccuratesegmentation,helpstoclarifythesemanticsofaparticularregion,and1.1
7、RelatedWorkenablesasimplequeryinginterfaceforregion-basedimageContent-basedimageretrievalisde nedastheretrievalofretrievalsystems.TheIRMhasbeenimplementedasapartrelevantimagesfromanimagedatabasebasedonauto-ofourexperimentalSIMPLIcityimageretrievalsystem.Thematica
8、llyderivedimageryfeatures.Theneedforecientapplicationtoadatabaseofabout200,000general-purposecontent-basedimageretrievalhasincreasedtremendouslyinimagesshowse
