資源描述:
《Illumination robust interest point detection.pdf》由會員上傳分享,免費在線閱讀,更多相關(guān)內(nèi)容在學(xué)術(shù)論文-天天文庫。
1、ComputerVisionandImageUnderstanding113(2009)565–571ContentslistsavailableatScienceDirectComputerVisionandImageUnderstandingjournalhomepage:www.elsevier.com/locate/cviuNoteIlluminationrobustinterestpointdetectionMuratGevrekci,BahadirK.Gunturk*DepartmentofElectri
2、calandComputerEngineering,LouisianaStateUniversity,BatonRouge,LA70803,UnitedStatesarticleinfoabstractArticlehistory:Mostinterestpointdetectionalgorithmsarehighlysensitivetoilluminationvariations.Thispaperpre-Received12February2008sentsamethodto?ndinterestpoints
3、robustlyevenunderlargenon-uniformphotometricchanges.Accepted26November2008Themethod,whichwecallilluminationrobustfeatureextractiontransform(IRFET),determinessalientAvailableonline24December2008interestpointsinanimagebycalculatingandanalyzingcontrastsignatures.A
4、contrastsignatureshowstheresponseofaninterestpointdetectorwithrespecttoasetofcontraststretchingfunctions.TheIRFETKeywords:isgenericandcanbeusedwithmostinterestpointdetectors.Inthispaper,wedemonstratethattheCornerdetectionIRFETimprovestherepeatabilityrateoftheHa
5、rriscornerdetectorsigni?cantly(byaround25%onaver-Interestpointdetectionageintheexperiments).IlluminationinvarianceHarriscornerdetectoró2008ElsevierInc.Allrightsreserved.1.Introductionpoint,bothmagnitudesshouldbelarge.Sincethecalculationofeigenvaluesandeigenvect
6、orsiscostly,acornerstrengthmeasureInterestpointdetectionisanecessarytaskinavarietyofcom-basedonthedeterminantandtraceoftheautocorrelationmatrixputervisionapplications,includingobjectrecognition,tracking,isproposed.Thelocalmaximaofthecornerstrengthmaparecho-ster
7、eoimaging,andmosaicconstruction.Althoughtheterminter-senasthecornerpoints.ThismethodisknownastheHarrisortheestpointdetectorissometimesinterchangeablyusedwiththePlesseycornerdetectorintheliteratureanditisprobablythemosttermcornerdetector;interestpointsdonotneces
8、sarilycorrespondpopularinterestpointdetector.ShiandTomasimethod[3]isbasedtophysicalcornerpoints.TheperformanceofaninterestpointontheHarriscornerdetectorbutusestheminimumofth