搜索
355Vol.35,No.520095ACTAAUTOMATICASINICAMay,2009WBCT112Contourlet(Wavelet-basedcon-tourlettransform,WBCT).BP,WBCT,5.,.,.,,BP,ContourletTP391.41AssessmentofIrisImageQualityBasedonWavelet-basedContourletTransformCHENRui1LINXi-Rong1DINGTian-Huai2AbstractWavelet-basedcontourlettransform(WBCT)cane±cientlyre°ectvisualcharacteristicsofiristextures.Afterlo-calizingaregionofevaluationbyusingacascadedBPneuralnetwork,theregionofimageisdecomposedwithWBCT.Then,5imagequalityassessmentindicesarede¯nedandcalculatedfordefocusedimages,motionblurredimages,contactlenswearimages,eyelashandeyelidoccludedimages,respectively.Exper-imentalresultsindicatethattheseindicescanpreciselyevaluatethesecaseswithlowcomputationalcost.Moreover,theevalu-ationresultshavegoodconsistencywithsubjectiveassessmentofhumanbeings.Comparedwiththeexistingmethods,ourmethodachievesbetterperformance.KeywordsImagequalityassessment,regionofevalua-tion,BPneuralnetwork,wavelet-basedcontourlettransform(WBCT),.,.Daugman[1].Zhang[2].Ma[3],,.[4],,.2008-05-272008-12-08ReceivedMay27,2008;inrevisedformDecember8,20081.5180552.1000841.GraduateSchoolatShenzhen,TsinghuaUniversity,Shenzhen5180552.StateKeyLaboratoryofPrecisionMeasurementTech-nologyandInstruments,TsinghuaUniversity,Beijing100084DOI:10.3724/SP.J.1004.2009.00618[5],.[6],,.[7],,.[8],.,[9](Humanvisualsystem,HVS)..,,.,.WBCT(Wavelet-basedcontourlettransform)[10],,,,,.WBCT,BP,WBCT,,.1WBCT2002,Do[11]:Contourlet.,,.Contourlet,LP(Laplacianpyramid),.2004,Eslami[10]:Contourlet(Wavelet-basedcontourlettransform,WBCT).WBCT:;(Directional¯lterbank,DFB)2n,.M£Nf(x;y),Wj;HLWj;LHWj;HH,.j,ªj;HL(n)ªj;LH(n)ªj;HH(n).3lj,kW(lj)j;HL;kW(lj)j;LH;kW(lj)j;HH;k(0·k2(lj)),Wj;i=2lj¡1©k=0W(lj)j;i;k;i=HL;LH;HH(1),©.´(lj)j;i;k(n)=Xm2zzz2g(lj)k(n¡S(lj)km)ªj;i(m)(2)5:WBCT619,g(lj)k,S(lj)k.lj,c(lj)j;k[n]=hf;´(lj)j;i;k(n)i(3)(1)»(3),WBCTDFB.\Haar,DFBMcClellan\9¡7[12].48,1.1WBCTFig.1WBCTdecompositionforirisimages2,.PALCCD.,.,..2.2Fig.2Severaltypesofpoorqualityirisimages2.1[13],,,.,.,QdefocusQdefocus=Xi2¡3£MXm=12¡4£NXn=1(ci(m;n)¡¹ci)22¡3£M£2¡4£N(4)¹ci=2¡3£MXm=12¡4£NXn=1ci(m;n)2¡3£M£2¡4£Ni=29;30;31;32;45;46;47;482.2,[14].,.,,QmotionQmotion=Xi2¡3£MXm=12¡4£NXn=1([ci(m;n)¡ci(m+1;n)]2+[ci(m;n)¡ci(m;n+1)]2);i=37;38;39;40(5)2.3,§45±[15].,.,,QeyelashQeyelash=Xi2¡3£MXm=12¡2£NXn=1(ci(m;n)¡¹ci)2Xj2¡2£MXm=12¡3£NXn=1(cj(m;n)¡¹cj)2(6),i=9;10;11;12;j=13;14;15;16.2.4,.,,QeyelidQeyelid=¯¯¯¯¯¯Xi2¡3£MXm=12¡2£NXn=1ci(m;n)¡Xj2¡2£MXm=12¡3£NXn=1cj(m;n)¯¯¯¯¯¯2¡2£M£2¡2£N(7),i=1;2;3;4;j=5;6;7;8.2.5,.,,.,.,QlensQlens=Xi2¡3£MXm=12¡2£NXn=1ln[1+ci(m;n)]22¡2£M£2¡2£N(8)62035,i=17;18;19;20;21;22;23;243,,,.,BP,.:,,;,,;,,,.BP,,Paul[16].,25£25.,,,,,..4,THUIris1.0CASIA-IrisV3[17].Matlab7,Petium2.0GHzCPU,1GRAM.:2800,300;800;800;800;800;200;200.,.,BP,18.6ms.4.1,.3,,,,.4.2,6.(1),.QdefocusQmotion,,.QeyelashQeyelid,,.Q1lensQ2lens,.,.3Fig.3Distributionsofqualityevaluationindices1Table1ComputationalresultsofqualityevaluationindicesNumberQdefocusQmotionQeyelashQeyelidQ1lensQ2lens1149944500.960.2531025282024801.180.3844070351611301.310.5253012043418301.980.6162515052545602.420.7786819061934902.510.91103022071184202.631.1311292708343403.131.5412372905:WBCT6212Table2TestingresultsandperformancecomparisonDaugman[1]Zhang[2]CCR(%)98.899.196.594.491.298.387.883.498.9¹t(ms)27.440.233.628.425.325.3122353324.3(Correctclassi¯-cutionrate,CCR)(¹t)2.,,,.Daugman[1]Zhang[2],,,.5WBCT.,WBCT,.,,.References1DaugmanJ.Howirisrecognitionworks.IEEETransactionsonCircuitsandSystemsforVideoTechnology,2004,14(1):21¡302ZhangGH,Salganico®M.MethodofMeasuringtheFocusofClose-UpImageofEyes,TechnicalReportU.S.Patent5953440,WhittendaleDriveMoorestown,USA,19993MaL,TanTN,WangYH,ZhangDX.Personalidenti-¯cationbasedoniristextureanalysis.IEEETransactionsonPatternAnalysisandMachineIntelligence,2003,25(12):1519¡15334WeiZS,TanTN,SunZN,CuiJL.Robustandfastassessmentofirisimagequality.In:ProceedingsoftheIn-ternationalConferenceonBiometrics.HongKong,China:Springer,2006.464¡4715KalkaND,ZuoJY,SchmidNA,CukicB.Imagequalityassessmentforirisbiometric.In:ProceedingsoftheInter-nationalSocietyforOpticalEngineering.Kissimmee,USA:SPIE,2006.1¡116ChenY,DassSC,JainAK.Localizedirisimagequalityusing2-Dwavelets.In:ProceedingsoftheInternationalCon-ferenceonBiometrics.HongKong,China:Springer,2006.373¡3817PanLL,XieM.Thealgorithmofirisimagequalityevalu-ation.In:ProceedingsoftheInternationalConferenceonCommunications,Circuits,andSystems.Kokura,Japan:IEEE,2007.616¡6198ChenJi,HuGuang-Shu,XuJin.Irisimagequalityevalua-tionmethodbasedonwaveletpacketdecomposition.Jour-nalofTsinghuaUniversity(ScienceandTechnology),2003,43(3):377¡380(,,..(),2003,43(3):377¡380)9LiZH,ZhangCH,MingX,ZhaoYH.Researchirisse-rialimagesqualityassessmentmethodbasedonHVS.In:ProceedingsofOpticalDesignandFabrication.Changchun,China:SPIE,2006.461¡46610EslamiR,RadhaH.Wavelet-basedcontourlettransformanditsapplicationtoimagecoding.In:ProceedingsoftheInter-nationalConferenceonImageProcessing.Singapore,Singa-pore:IEEE,2004.3189¡319211DoMN,VetterliM.Thecontourlettransform:ane±-cientdirectionalmultiresolutioni