effective extraction of gabor features for adaptive mammogram retrieval

《effective extraction of gabor features for adaptive mammogram retrieval》由会员上传分享，免费在线阅读，更多相关内容在学术论文-天天文库。

1、EFFECTIVEEXTRACTIONOFGABORFEATURESFORADAPTIVEMAMMOGRAMRETRIEVALChia-HungWei,YueLi,Chang-TsunLiDepartmentofComputerScience,UniversityofWarwick,UK{rogerwei,yxl,ctli}@dcs.warwick.ac.ukABSTRACT1)AGabor-filteringmethodfordescribingtexturalfeaturesisproposed,whichappliesthephysicalpropertiesofaprob-

2、Breastcancerisoneofthemostcommondiseasesamongabilitywavetoprobabilitytransformation,andcomputeswomen.Content-basedmammogramretrievalhasbeenpro-texturalfeatures;posedtoaidvariousmedicalprocedures.Todevelopacon-2)Anadaptivestrategyforfeatureandfilterselection,andtent-basedmammogramretrievalsyste

3、m,texturalfeaturefeatureweighting,isproposed,whichutilizesauser’srele-extractionisoneofthecrucialrequirements.Thisstudyvancefeedbacktoreducetheredundancyintherepresenta-proposesaGaborfilteringmethodfortheextractionoftex-tionandincorporatetheuser’sinformationneedsinimageturalfeatures,whichfirst

4、lyperformsGaborfilteringontheretrieval;underlyingimage,appliesthephysicalpropertiesofaprob-3)Integratingbothschemes(theGabor-filteringmethodandabilitywavetoprobabilitytransformationandthencom-theadaptivestrategy)forcontent-basedmammogramre-putesfeaturestodescribethetexturalpatternofthemam-trie

5、valissuggested,asshowninFigure1.mogram.Thisstudyalsoproposesanadaptivestrategyforfeatureselection,filterselectionandfeatureweighting,GaborProbabilityFeaturewhichutilizesauser’srelevancefeedbacktoreducethere-FilteringTransformationCalculationdundancyintherepresentationandincorporatestheuser’sFe

6、atureExtractioninformationneedsinimageretrieval.Experimentalresultsshowthathypothesistestscaneffectivelyfinddiscriminatedfeaturesandthisretrievalsystemcanimproveitsperform-Featureancethroughjustafewroundsofrelevancefeedback.Mammo-FeaturegramsExtractionImageDatabase(with1(Filter.FeatureDescript

7、ors)..FeatureOff-lineFeatureExtraction1,…,FilterOn-lineImageRetrievalm1.INTRODUCTION(FilterQueryFeaturenExampleExtractionSimilarity)Measure1,…,FilterMammographyhasproventobeareliablemethodforde-(Euclideantectingbreastcancersoanenormousn