资源描述：
《Pigment Melanin- Pattern for Iris Recognition》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、1PigmentMelanin:PatternforIrisRecognitionMahdiS.Hosseini,BabakN.Araabi,andHamidSoltanian-ZadehAbstractRecognitionofirisbasedonVisibleLight(VL)imagingisadicultproblembecauseofthelightre
ectionfromthecornea.Nonetheless,pigmentmelaninprovidesarichfeaturesourceinVL,unavailab
2、leinNear-Infrared(NIR)imaging.Thisisduetobiologicalspectroscopyofeumelanin,achemicalnotstimulatedinNIR.Inthiscase,aplausiblesolutiontoobservesuchpatternsmaybeprovidedbyanadaptiveprocedureusingavariationaltechniqueontheimagehistogram.Todescribethepatterns,ashapeanalysismet
3、hodisusedtoderivefeature-codeforeachsubject.Animportantquestionishowmuchthemelaninpatterns,extractedfromVL,areindependentofiristextureinNIR.Withthisquestioninmind,thepresentinvestigationproposesfusionoffeaturesextractedfromNIRandVLtoboosttherecognitionperformance.Wehaveco
4、llectedourowndatabase(UTIRIS)consistingofbothNIRandVLimagesof158eyesof79individuals.Thisinvestigationdemonstratesthattheproposedalgorithmishighlysensitivetothepatternsofcromophoresandimprovestheirisrecognitionrate.IndexTermsIrisBiometrics,Visible-Light(VL),Near-Infrared(N
5、IR),PigmentMelanin,Eumelanin,ShapeAnalysis,ImageEnhancement,Regularized(Tikhonov)FilteringandVariationalBinarization.I.IntroductionRISrecognitionisoneofthemostreliablenon-invasivemethodsofpersonalidenti cationowingItothestabilityoftheirisoverone'slifetime.Pioneerworkoniri
6、srecognition{asthebasisofmanycommercialsystems{wascarriedoutbyDaugman[1].Inthisalgorithm,2DGabor ltersareadoptedtoextractoriented-basedtexturefeaturescorrespondingtoagivenirisimage.AfterDaugman,otherresearchershavecontributednewmethodstoarriveatalternativealgorithmswithlo
7、wcomputationalburden,lessSNRandmorecompactcodes,e.g.[2],[3],[4],[5],[6]and[7].Mostfeatureextractionmethodshavebeenimplementedthroughmulti-resolutionalanalysis,e.g.applyingLaplacianpyramidconstructionwithfourdi erentresolutionlevels[2];zero-crossingrepresentationof1Dwavele
8、ttransformatvariousresolutionlevelsofavirtualcircle[3];2Dwaveletdecomposition[4];and1DDiscreteCo
