An overview of touchless 2D fingerprint recognition

Touchless fingerprint recognition represents a rapidly growing field of research which has been studied for more than a decade. Skiptomaincontent Advertisement SearchallSpringerOpenarticles Search Anoverviewoftouchless2Dfingerprintrecognition DownloadPDF DownloadPDF Review OpenAccess Published:24February2021 Anoverviewoftouchless2Dfingerprintrecognition JannisPriesnitz  ORCID:orcid.org/0000-0002-0985-77351,ChristianRathgeb1,NicolasBuchmann2,ChristophBusch1&MarianMargraf2  EURASIPJournalonImageandVideoProcessing volume 2021,Article number: 8(2021) Citethisarticle 8242Accesses 12Citations 8Altmetric Metricsdetails AbstractTouchlessfingerprintrecognitionrepresentsarapidlygrowingfieldofresearchwhichhasbeenstudiedformorethanadecade.Throughatouchlessacquisitionprocess,manyissuesoftouch-basedsystemsarecircumvented,e.g.,thepresenceoflatentfingerprintsordistortionscausedbypressingfingersonasensorsurface.However,touchlessfingerprintrecognitionsystemsrevealnewchallenges.Inparticular,areliabledetectionandfocusingofapresentedfingeraswellasanappropriatepreprocessingoftheacquiredfingerimagerepresentthemostcrucialtasks.Also,furtherissues,e.g.,interoperabilitybetweentouchlessandtouch-basedfingerprintsorpresentationattackdetection,arecurrentlyinvestigatedbydifferentresearchgroups.Manyworkshavebeenproposedsofartoputtouchlessfingerprintrecognitionintopractice.Publishedapproachesrangefromselfidentificationscenarioswithcommoditydevices,e.g.,smartphones,tohighperformanceon-the-movedeploymentspavingthewayfornewfingerprintrecognitionapplicationscenarios.Thisworksummarizesthestate-of-the-artinthefieldoftouchless2Dfingerprintrecognitionateachstageoftherecognitionprocess.Additionally,technicalconsiderationsandtrade-offsofthepresentedmethodsarediscussedalongwithopenissuesandchallenges.Anoverviewofavailableresearchresourcescompletesthework. IntroductionFingerprints,i.e.,ridgeandvalleypatternsonthetipofahumanfinger,areoneofthemostimportantbiometriccharacteristicsduetotheirknownuniquenessandpersistenceproperties[1,2].Automatedtouch-basedfingerprintrecognitionhasbeenatopicofresearchforseveraldecades[3].Nowadays,large-scaletouch-basedfingerprintrecognitionsystemsarenotonlyusedworldwidebylawenforcementandforensicagencies,buttheyarealsodeployedinthemobilemarketandinnation-wideapplications[2,4].However,thetouch-basedfingerprintcapturingprocesssuffersfromdistinctproblems,e.g.,signalsoflowcontrastcausedbydirtorhumidityonthesensorplate,latentfingerprintsofprevioususers,ordistortionsduetoelasticdeformationofthefingercausedbythepressurewhichisputonthesensorplate[5].Inaddition,aninconvenientacquisitionprocessandhygienicconcernsmaylowertheuseracceptabilityoftouch-basedfingerprintsystemsandhence,limittheirdeployment.Totackletheseshortcomingsoftouch-basedfingerprintrecognitionsystems,thefirsttouchless(alsoreferredtoascontactless)fingerprintrecognitionschemewasproposedbySongetal.in2004[6].Sincethen,aconstantlygrowingnumberofcontributionsrelatedtothistopichavebeenpublishedeachyearbynumerousresearchlaboratoriesworkinginthefieldofbiometrics,asillustratedinFig. 1.Conceptualadvantageslikealessconstrainedacquisitionprocesspavethewayfornewapplications,improvesusabilityandhence,useracceptance.Further,fingerimagesacquiredbyatouchlesssensorexhibitnodeformationandcomprisenolatentfingerprints.Thesemajoradvantagesmotivatedalargeamountofworkspublishedinrecentyears. Fig.1Yearlyamountofpublications.Amountofpublicationsinmajorconferencesorjournalssince2004dedicatedtothetopicoftouchlessfingerprintrecognitionFullsizeimageThisworkaimsatprovidingacomprehensiveoverviewofpublishedscientificliteratureinthefieldoftouchlessfingerprintrecognition.Itisnotintendedtore-evaluateproposedapproachesasimplementationsofmanyworksarenotpubliclyavailableandre-implementationsmightlackimportantoptimizationsorrequirespecificsensorhardware.Moreover,fortechnicaldetailsofsurveyedapproaches,thereaderisreferredtotheaccordingpublications.Wherepossible,resultsofpublishedworksarepresentedinacomparativemanner.Ifauthorsprovidedasingleresultinthepublicationtext(e.g.,intheabstractorsummary),thosevaluesaretakendirectly.Otherwise,arepresentativeresultischoseningoodfaithfromthepresentedplotsandtables.Whiletouchlessfingerprintrecognitiontechnologieshavebeeninvestigatedforsomeyears,thecorrespondingliteratureisdispersedacrossdifferentpublicationmediaandoverviewworksmostlyfocusonspecificprocessmodules.ParzialeandChen[7]elaboratedonthedifferencesof2Dand3Dacquisitiontechnologies,processingstrategies,andqualityaspects.Further,theauthorsgaveanoverviewonpresentationattackdetection(PAD)schemes.KhalilandWan[8]reviewedstate-of-the-artalgorithmsalongthepreprocessingpipelineandaddressPAD.Eventhough,theirworkhighlightssomeimportantissuesinthefielditlacksacomprehensivediscussionofcurrentapproaches.Labatietal.[5]conductedacomparativeoverviewof2Dversus3Dtouchlessfingerprintrecognitionandaddresstheprocessingoftouchlessfingerprintstotouch-basedequivalentfingerprintsusingunwrappingalgorithms.Moreover,theauthorsprovideahigh-leveldiscussionofdifferentfeatureextractionandcomparisonsubsystems.AbriefsurveyofmobiletouchlessfingerprintrecognitionusingsmartphonesascapturingdevicehavebeenpresentedbyMalhotraetal.[9].Mil’shteinandPillai[10]presentashortcomparativereviewoftouchlessandtouch-basedschemesaswellasaselectivesummaryofstate-of-the-arttouchlessacquisitiontechniques.Inaddition,theauthorsbrieflydiscusschallengesoftouchlessrecognition.Labatietal.[11]providedamoreelaboratedoverviewofthewholerecognitionpipelinewhichiscompletedbyadiscussionoflivenessdetectionalgorithms,nonidealitiesofcurrentapproaches,andaperformancesummary.Aspreviouslymentioned,thepublishedoverviewpapersaremostlyrestrictedtoparticularsubsetsofthetopic,i.e.,subsystemsofatouchlessfingerprintrecognitionsystem.Asthefactthattheexistingsurveysareeithernotcomprehensiveoroutdated,thisworkaimsatprovidingamorecompleteoverviewofthestate-of-the-artoftouchless2Dfingerprintrecognition.Thefirstpartisstructuredaccordingtothepipelineofatouchlessfingerprintrecognitionsystem.Itprovidesthereaderbriefoverviewofmainprocessingsteps,aswellasadetailedsummaryofproposedapproaches.Inasecondpart,anin-depthdiscussionofissuesandchallengesisprovided.Furthermore,availableresearchresourcesaredescribedindetail.Thissummaryprimarilyaddressesbiometricresearchersandpractitionersaimingtogainanoverviewofthecurrentstate-of-the-artofthetopic.Apartfromthestandardizedtermsanddefinitions[12],thefollowingtaxonomywillbeusedthroughoutthiswork: Fingerimageorfingerphotoreferstoanimageacquiredusingatouchlesscapturedevice,e.g.,smartphonecamera,whichcontainsoneormorefingersofasubject. Fingerprintimagereferstoafingerimagecroppedtoanarearepresentingafingerprint,i.e.,fingertips. Fingerprintreferstoapreprocessedtouchlessfingerprintimageorafingerprintcapturedbyatouch-basedsensor. Furthermore,adistinctionismadebetweenthecapturingofafingerimagewithoutanypreprocessingandtheacquisitionofafingerprintimagewhichincludesanenhancementbysomepreprocessingalgorithms.ItshouldbenotedthattheISO/IEC2382Part37standardsuggeststheusageofthetermcapturingprocess[12].ThegeneralbiometricworkflowofatouchlessfingerprintrecognitionsystemissketchedinFig. 2.Thefirstpartofthisworkisstructuredaccordingly:Section2describesdifferentfingerimagecapturingapproaches.InSection3,theprocessingstepswhicharenecessarytoachieveahigh-qualitybiometricsamplearedescribed.Section4highlightstouchlessqualityassessmentfollowedbyasummaryoffeatureextractionandcomparisonapproachesinSection5andSection6.ThesecondpartdiscussesdifferentissuesandchallengesinSection7.AnoverviewontouchlessbiometricdatabasesisfurthergiveninSection8.Section9finallydrawsaconclusion. Fig.2Modulesoffingerprintrecognition.Overviewonthemainmodules(sub-systems)ofagenerictouchlessfingerprintrecognitionsystemFullsizeimageCapturingprocessDuringatouchlesscapturingprocess,oneormorefingersarepresentedtoanopticalcapturingdevice.Thesedevicescaneitherbeprototypicalhardwaredesignsassembledbytheresearchersorgeneralpurposedeviceswhichareadaptedtothespecialneedsoftouchlessfingerprintrecognition.TheNationalInstituteofStandardsandTechnology(NIST)[13]publishedaguidancedocumentfortheevaluationoftouchlessfingerprintcapturing.Thedocumentaccuratelydefinesrequirementsfortheassemblyoftouchlessfingerprintcapturingdeviceswithrespecttodifferentapplicationscenarios.Figure 3depictsimpressionsofafingerprintcapturedwithatouch-basedfingerprintsensor(Fig. 3a)andathecorrespondingfingerimageacquiredusingatouchlessdevice(Fig. 3b).Itisobservablethatthetouch-basedfingerprintcanbedirectlyusedforfeatureextractionwhereasthecorrespondingtouchlessfingerprintimagerequiresfurtherpreprocessing. Fig.3Twoimpressionsofthesamefinger:atouch-basedfingerprintacquiredwithaCrossmatchGuardian200;btouchlessfingerprintimagecapturedwithaSamsungGalaxyS8.BothimagesaremanuallycuttorepresentonlythefingerprintareaFullsizeimagePrototypicalhardwaredesignManyprototypicalhardwaredesignsrelyonelaboratedcapturingtechnologiesadoptedfromotherresearchareastoobtainfingerimagesofhighquality.Table 1listsmostrelevantworkscategorizedbyapproachandorderedbytheyearofpublication.Alllistedapproachesfocusonovercomingknownchallengesoftouchlessfingerprintcapturinglikeunconstrainedenvironmentalinfluences,thelackofdeformations,orfocusingissues. Table1OverviewofmostrelevantprototypicalhardwaresetupsforcapturingfingerimagesFullsizetableSeveralauthorscombineabox-likesetupwithLEDstoachieveapredicableilluminationandtoexcludeenvironmentalinfluences[15,16,18].LEDarrangementsaroundthefingerleadtoahomogeneouscontrastonthefingerprintarea.Coloredilluminationcanalsoemphasizethefingerprintcharacteristicsandhenceleadtoimprovedresults[16].Themajorityofcapturingsetupsusedfingerguidanceinformofcircularholes[16]orfixedfingerplacements[20].Tsaietal.[17]presentedamoreunconstrainedapproachwhichworkswithoutaboxandfingerguidance.Theauthorsusedastrongilluminationcombinedwithasmalldistancebetweenthelensandthefingertiptominimizeenvironmentallights.Avariable-focusliquidlenswasabletoacquirehigh-qualityfingerimagesofmovingfingers.Toovercometheissueoffingerprintdistortions,Palmaetal.[20]andMil’shteinetal.[14]presentedcapturingdevicesusingrotatinglinescancameras.Theacquiredfingerimageslicesweremergedtogethertoanail-to-nailrolledfingerprintimage.Thisimpressionhassignificantlyfewerdistortionsthanatouch-basedfingerprint.Alternatively,Wangetal.[15]suggestedasetupofthreecamerasarrangedaroundthefingertiptoacquirefingerphotosofdifferentorientationwhicharestitchedtogether.Acontinuousimageanalysisassessedifthefingerwaspositionedproperlyandenabledaconvenientcapturingofhigh-qualityfingerimages.Mil’shteinetal.[14]andRamachandraetal.[18]showedthepossibilityofcombiningthecapturingoffingerprintsandfingerveinsinmulti-modaldevices.Ramachandraetal.[18]usedlow-costequipmentsuchasanindustrialcamerawithamonochromesensor.Weissenfeldetal.[19]introducedamobilehand-helddevicewhichcapturedfaceandfingerimagesusingasinglesensor.GeneralpurposedevicesIncontrasttoelaboratedhardwaresetups,manyresearchgroupsusegeneralpurposedevicestocapturefingerimages.MostrelevantapproachesaresummarizedinTable 2sortedbytypeofrecordingdevice. Table2OverviewofmajorcontributionsusinggeneralpurposedevicesforcapturingFullsizetableFirstexperimentsongeneralpurposedeviceswereconductedbyLeeetal.[21]whousedthecameraofamobilephonewithanexternalLEDlighttoacquirefingerimages.Hiewetal.[36]alsousedanexternalilluminationalongwithasemi-professionalcamerainaboxsetup.Inbothschemes,thefingerimageswereacquiredcompletelymanual.Severalearlyworksinvestigatedtheapplicabilityofwebcamsforfingerimageacquisition.Majoradvantagesareaffordablepriceandaneasyconnectivitytoacomputer[24,26,27].Allcontributionsusedamanualcapturingprocessandnoadditionalillumination.Additionally,PiuriandScotty[24]conductedanexperimentwithexternalilluminationbutwerenotabletoachievesignificantperformancebenefits.Nevertheless,theauthorsreportedaccurateresultsinatouchlesstotouch-basedinteroperabilityscenario.Itisworthnotingthatdespitetheratherlowimagequalityofwebcams,abiometricrecognitionscenariocouldbeestablishedwithsuchdevices[26]usinglevel-0features.Level-0featurestypicallyrefertolocaltexturepatternslikelinestructuresordominantlocalorientations.Nowadays,smartphonesaremostoftenusedforcapturingbecausetheyarewidelyavailable,havehigh-qualitycameras,andcanprovideimmediateuserfeedback.Here,themostpromisingsettingsaretokeeptheauto-focusactivatedandifavailableusethemacromode.Additionally,theflashshouldbeenabled[29,37].ExternalextensionslikeadditionallightsandmacrolensesareconsideredasbeneficialbySagirogluetal.[38].Severalauthorssuggestedusingon-screenfingerguidanceforahighuserconvenienceandaneasierfingerprintprocessingworkflow[29,33,35].Herethecameraviewpresentedonthescreeniscombinedwithalinerepresentingthefingercontour.Modernsmartphonesareabletoprocessandqualifyvideostreamsinordertoselecttheframewhichcontainsafingerimageofhighquality[30,39].AconvenientautomaticcapturingcomparabletotheapproachofWangetal.[15]canbeestablished.Moreover,Carneyetal.[33]andWeissenfeldetal.[40]proposedthecapturingofawholeslaphandinoneimagewhichmakesthecapturingofuptofourfingerprintsmoreconvenient.Severalworksconsideredfingerimagecapturingunderdifferentenvironmentalinfluences[32,41–43].Theauthorsconcluded,thatthecapturingitselfisnotlimitedbydifferentlightsituationsorindoorandoutdoorenvironments.Nevertheless,varyingbackgroundsmighthaveamajorinfluenceonfurtherprocessing.Duetothehugevarietyofsmartphones,severalworksinvestigatedoninteroperabilitybetweendifferentmodels[28,34,41].Itisobservablethattherearenohugeperformancedifferencesbetweenparticularmodelsofthesamegeneration.Debetal.[34]alsoshowedthatfingerprintimagesacquiredbylow-costsmartphonescouldbecomparedtotouch-basedfingerprints.Thetestedcommercialappsshowedapracticalbiometricperformance.Anail-to-nailrolledequivalenttouchlessfingerimageisadesirablegoaltoachievealargeregionofinterest(ROI).Alkhathamietal.[31]proposedanail-to-nailrolledfingerimagebymosaickingthreeimagesacquiredsequentiallywithonesmartphone.Duringthecapturing,thesubjectwasaskedtoperformavirtualrollingofhisfinger.Allthreeimageswerestitchedtogethertoformalargerfingerprint.Level-3characteristics,i.e.,sweatpores,ontouchlessimagedatawerefirstlyanalyzedbyGenoveseetal.[23].Theauthorsusedanoff-the-shelfcameraandagreenLEDillumination.Inaconstrainedsetupwithfixeddistancebetweenfingerandsensor,theauthorscapturedaccuratefingerimageswitharesolutionof≈3800ppiwhichissufficientforextractinglevel-3featureswhichrefertosweatpores.PreprocessingpipelineThecapturedimagedatadiffersfundamentallybetweentouchlessandtouch-basedacquisitiondevices.Mosttouch-basedschemesproduceagray-scaleimageinwhichtheridgeskinareatouchingthescannerssurfaceisshowninblack(ordarkgrayvalues)whilevalleyandbackgroundareaiswhite(orlightgrayvalues).Ingeneral,thesesamplesareuseddirectlyforfeatureextractionwithoutextensivepreprocessing.Themajorityoftouchlessfingerimageacquisitionschemesdelivercolorimageswhichrequireacomprehensivepreprocessingpriortotheextractionoffeatures.Basicchallengesarealowridgevalleycontrast,ablurredROI,andadisplaced,rotated,orpitchedfinger.Further,principallydifferentappearances,e.g.,thelackofskindeformation,causeincompatibilities.Theimageprocessingpipelinehastobedevelopeddependentontheselecteddeviceandtheobservedenvironmentalcircumstancesduringthecapturing.Foranexamplefingerimage,atouchlesspreprocessingpipelineisillustratedinFig. 4.Inrecentyears,touchlessfingerimagepreprocessingevolvedtoaheterogeneoustopicofresearchwithmanydifferentapproachesandcontributors.Unfortunately,thefieldlacksaharmonizedvocabularyinordertocomparedifferentapproaches.Togetaclearunderstandingofthepreprocessingsteps,wedefinefrequentlyusedtermsasfollows: 1. Fingerdetection:intheinitialstep,oneormorefingersaredetected(orsegmented),e.g.,basedoncolororshapeanalysis,seeFig. 4a–c. Fig.4Touchlesspreprocessingworkflow.ExampleofatouchlesspreprocessingworkflowbasedonafingerimagemanuallytakenbyaSamsungGalaxyS8Fullsizeimage 2. Grayscaleconversion,ROIextraction,andorientationestimation:thefingerimageisconvertedtograyscaleanddetectedfingersarefurthercroppedtoextractfingerprintimageswhicharealignedforfurtherprocessing,seeFig. 4d. 3. Fingerprintimageenhancement:generalimageprocessingtechniquesareemployedtoimprovethecapturedfingerimage,i.e.,increasecontrastandsharpness,seeFig. 4e. 4. Furtherpreprocessing:thefingerimageisenhancedtoobtainfingerprintsandtopronouncetheirfeatures,e.g.,byskeletonizing,seeFig. 4f,g.Theseapproachescanbedirectlytakenfromthetouch-baseddomainandarenotdiscussedindetailinthiswork. In2012,KhalilandWan[8]presentedasurveyonthespecialtopicofpreprocessingfingerimagesacquiredwithmobilephones.Theauthorshighlightedtherelevanceofthisfieldofresearchandsummarizedthedifferencesbetweenthetouchlessandthetouch-baseddomain.Elaboratedpreprocessingworkflowshavetobedevelopedespeciallyforcommoditydevicesinordertocompensatethelimitedcapabilitiesofbuilt-incamerasandenvironmentalsideeffects.Thefollowingsubsectionssummarizeproposedapproachesforeachprocessingstage.Table 3additionallyhighlightsfundamentalchallengesofprocessingtouchlessfingerimagesandlistssuggestedmethodstoovercomethesechallenges. Table3OverviewofchallengesduringthepreprocessingoffingerimagesandproposedapproachesFullsizetableFingerdetectionandsegmentationUnconstrainedcapturingsystems,whichdonothaveafingerguidancebasedondedicatedhardwareoranon-screenguidance,requireafingerdetection.Suchanalgorithmdetectsthepositionandorientationofthefingerandformsthebasisforanautomaticcapturingsystem.Theimageisthensegmentedandcuttothefingerprintcontainingarea.Fourdifferentapproachescanbedistinguished,whereasinpracticeimplementationsoftenapplyacombinationofthem: Sharpness:Sharpness-basedapproachesexploitthedifferencebetweenthefocusedsharpfingerareaandtheblurredbackground.Thiseffectismostsuitableonimagesacquiredwithaverysmallfinger-to-sensordistanceandawideopenaperture.TheearlyworkofLeeetal.[49]presentedafixedfocusreal-timescheme,whichselectedthebestfocusedandorientedimageoutofaseries.Theauthorsinvestigatedonthesuitabilityofgeneralpurposefocusmeasuringalgorithms.TheirexperimentshowedthattheVariance-Modified-LaplacianofGaussian(VMLOG)algorithmisbestsuitedforthetouchlessfingerprintcapturingdevicetheyused.Theauthorsalsocomparedafingermovingmethodwithafixedlenstoalens-movingmethodwithafixeddistancebetweensensorandfinger.Theyconcludedthattheformermethodispreferablewhichisquestionablefromtoday’sperspective.Asubsequentworkbythesameauthors[21]comparedthreesegmentationapproaches.Oneofthemwassharpness-basedandusedtheTenengradmethod[50]inthefrequencydomain.Here,aSobeloperatorwasusedtocalculatethehorizontalandverticalgradientsintheimage.Acertainthresholdwasestablishedtoseparatethesharpforegroundfromthebackgroundarea.Leeetal.[51]aimedatselectingthebestfocusedimageoutofavideostream.TheauthorsproposedanalgorithmbasedonaGaussianfiltertosegmentthesharpregionsofanimagewhichcorrespondedtothefingerregion. Shape:Theshapeofafingerishighlycommonforallfingerpositioncodes(i.e.,variousfingerinstancesfromthumbfingertolittlefinger),whichenablesadetectionviashape.Jonietzetal.[52]proposedaconjunctionofashape-andcolor-basedfingerdetectionusingedgepairing.Theauthorsappliedmachinelearning-basedalgorithmstothebinarizedimageintheLUVcolormodel.TheyalsousedHistogramofOrientedGradient(HOG)featureswithrichfeaturedescriptorsasbaselineandcomparedtheirresultswiththem. Contrastandcolor:Especially,ifacertainilluminationisused,adeterminationbasedonthecontrastorcolorisanefficientmechanismforfingerdetection.BasedonfindingsofHiewetal.[53]forthesegmentationinskinandbackgroundarea,ananalysisoftheYCbCrcolorspacerepresentsthemostpromisingapproach.Theresultisabinaryimagewithaseparationbetweenfingerimageareaandbackground.Theaboveapproachiswidelyadopted,modifiedtomeetdifferentprerequisites,andfurtherinvestigatedbymanyauthors[37,46,54,55].RaviandSivanath[27]showedthatextendingtheCrcomponentwithinformationoftheHSVandnRGBcolorspaceenablesapreciseisolationofafinger.Theauthorsusedacertainthresholdforeverycolorchannelandmergedtheresults.Wangetal.[44]presentedcomprehensiveresearchondifferentfingerilluminationsandcolormodels.Forthisreason,theauthorscapturedimageswithgreen,red,andblueilluminationandcomparedtheYCbCrcolormodelwithYIQandHSV.Alternatively,othercolormodelssuchasCMYK(magentachannel)[9]andCIELAB[39]werealsoinvestigated.Thisapproachwasadoptedinmanyotherpreprocessingworkflowssimilarto[37,46,55].Becauseofprerequisitesduringthecapturingprocess,mostapproachesconsideredonlythelargestsegmentedareaasfingerprint[37,55].Thecolor-basedsegmentationisoftencombinedwithanadaptivethresholding,e.g.,basedonOtsuimagethresholding[9,44,46,53].Hieretal.[53]alsodeterminedthemeanandcovarianceontheCbCrchannelstoimprovethesegmentationaccuracy.AnotherapproachbyLeeetal.[21]exploitedskincolorpropertieswithhelpofguidedmachinelearning.Thisapproachwasshowntorevealcompetitiveresultsbutismorecomplexcomparedtoothers.Asasecondscheme,theauthorssuggestedaregiongrowingapproach.Usinganinitialseedandasimilaritymeasurewithacertainthresholdthetestedpixelswereaddedtotheseed.ThisapproachisalsosuitableforROIextraction.WiththemeanshiftsegmentationRamachandraetal.[41]proposedanothercontrast-basedapproach.Thealgorithmfilterstheinputimageinthespatialdomainandsegmentsitbyfusingtheconvergencepointsinhomogeneousregions.Withthiselaboratedapproach,theauthorswereabletoachieveaccurateresultsinchallengingenvironments.Priesnitzetal.[56]presentedadeeplearning-basedsemanticsegmentationschemeforthehandareaaswellasfingertips.Theauthorsusedageneralpurposehandgesturedatasettotesttheiralgorithmagainstacolor-basedbaselinesegmentationalgorithm.Theproposedmethodshowedaccurateresultsespeciallyinchallengingenvironmentalconditions.Itshouldbecriticallynotedthatnoneofthediscussedapproachesconductedawideranalysisondifferentskincolortypes,e.g.,asdefinedin[57]. Imagedepthinformation:JonietzandJivet[58]presentedasegmentationapproachusingtheinformationofadepthsensorcombinedwithanRGBimagecapturedbysmartphones.Theauthorswereabletoextracttheslaphandfromabusybackgroundandproposedfurtherprocessing.Exploitingtheimages’depthinformationthesystemworkedespeciallywellinthepresenceofobjectsofsimilarcolor,e.g.,whentwohandswereplacedontopofeachother. ROIextraction,orientationestimation,andcorepointdetectionOnceafingerisdetected,theROIhastobeextractedwhichincludesthenormalizationtoaproperwidth,height,andresolution.Thispreprocessingstageassumesanextractedfingerimageasinput.Itshouldbenotedthat,especiallyinmoreconstrainedsetups,fingerdetectionandROIextractionisdoneinonestep[41].Intheirwork,PiuriandScotti[24]simplifiedthecolor-basedsegmentationapproachofLeeetal.[21]forROIextraction.Theauthorscombinedthisapproachwithafrequencyestimationmap.Moreover,theyusedaGaussianprobabilitydensityfunctionandperformedaregiongrowinginordertoextracttheROI.AcomparableapproachbyHiewetal.[53]exploitedtheridgelinecharacteristicsofthefingertip.Here,thesegmentedfingerwasdividedinnon-overlappingblocks.Ifaridge-linecharacteristicwasobservablewithinablock,itwasaddedtotheROI.Ramachandraetal.[41]alsoshowthatinconstrainedsetupsaROIextractionbasedonfingergeometrypropertiesisalsopossible.TheauthorscomputedtheROIstaticallybydetectingcharacteristicpointslikethefingertipanddiscontinuities.Sincemostfeatureextractorsarenotinvarianttotherotation,allfingerimagesmusthavethesameorientation.DongjaeLeeetal.[51]presentedarollingandpitchingestimationbycalculatingthedistancebetweenthecorepointandtheborderofthefingertip.Leeetal.[21]estimatedtheorientationbyiterativelycomputingtherobustregressionmethod.TheschemeusedtheSobeloperatoronsub-blocksoftheinputimagetocomputetheorientationofthelocalgradients.Asimpletechniqueonsegmentedfingerimagesistoapproximateatangentalongtheborderbetweenfingerandbackgroundandrotatetheimagetoapredefinedorientation[29].Incontrasttotheaforementionedcontributions,Ramachandraetal.[41]proposedapreprocessingpipelinewithoutarotationstageincombinationwitharotationinvariantfeatureextractor.Sisodiaetal.[55]alsointroducedanapproachwhichrotatesminutiaefeatures.Here,aminutiawhichisaboveapredefinedcorrelationthresholdhadtobedeterminedintheprobeandreferenceimages.Togetherwiththecorepointsofbothimages,arotationanglewascomputed.Regardinganapplicationtolargescaledatabases,theperformanceofthisapproachisquestionable.ManycomparisonalgorithmsrequireacorepointoraPrincipalSingularPoint(PSP)asreferencepoint.Severalworksusedtheridgelineorientationandcurvaturefordetectionofthecorepoint[53,55].Labatietal.[47]suggestedarathercomplexapproachwhichestimatesallsingularpointsfromtheglobalridgestructureusingcomputationalintelligenceclassificationtechniques.Leeetal.[51]usedthePoincaréindexfromthetouch-baseddomaindescribedin[59]toroughlydeterminethecorepoint.FingerprintimageenhancementAftertheextractionoftheROI,ridgelinecharacteristicshavetobefurtheremphasizedtoextractfeaturesaccurately.Simpleapproachesonlyadaptfingerprintimageswithkernelbasedoperationsinthespatialdomain[53],whereasmoreelaboratedalgorithmsexploitcombinationsofdifferentfiltersinthefrequencydomain[24].Fingerimageenhancementshouldresultinafingerprintimagewhichhasahomogeneousillumination.Anormalizationusingmeanandvariancefilters[53]orhistogramenhancementslikeContrastLimitedAdaptiveHistogramEqualization(CLAHE)[46,60]werefoundtobewell-suitedforthistask.Malhotraetal.[9]alsosuggestedtheanalysisofLocalBinaryPatterns(LBP)ontheridge-valeycontrastforenhancement.Moreover,Wasniketal.[39]suggestedaFrangiFilterwhichsearchesfortubularstructures.Animportantissueisthereductionofblurinthesourceimage.Toensurethis,PiuriandScotti[24]proposedacombinationoftheLucy-RichardsonandtheWienerfilter.Inaddition,theysuggestedablinddeconvolutionmethodtoenhanceimageswhichcouldnotbehandledbythealgorithmsproposedpreviously.Liuetal.[60]combinednoiseremovalandilluminationcorrection,andhistogramequalizationinspatialdomainwitharidgelinefrequencyestimationbasedonGaborfilters.Additionally,acontext-basedcorrectionissuggestedtoemphasizetheridge-linestructureonlowreliabilityareas.Thisapproachcomparesblocks(patches)ofthefingerprintwithadirectoryandsubstitutestheseblockswithmoreaccuratedata.Birajadaretal.[37]alsoexploitedphasecongruencyprocessinginthefrequencydomain.Theauthorsusethemonogenicextensionofareal2Dlog-Gaborisotropicwaveletfortheenhancement.Alaterworkofthesameauthors[35]confirmedthatthealgorithmalsoworksonalargescaledatasetcapturedinanunconstrainedenvironment.SimilarworkbasedontheaforementionedschemewaspresentedbySagirogluetal.[38].FurtherpreprocessingSpecialcapturingschemesorfeatureextractorsrequireadditionalpreprocessingsteps.Imagemosaickingorimagefusiondescribescompositionoftwoormoreimagestoonelargerfingerimage.Inthebestcase,thefusedimageexhibitsalargerROIandabetterimagequality.Mosaickingtechniquesbecameessentialinuse-caseswherealarge-sizedsensorisnotavailablebutarolledfingershouldbecaptured.IntheworksofChoietal.[61]andLiuetal.[62],theauthorsshowedcommonusecasesofmosaickingtouchlessimages.Three(virtual)imageswerestitchedtogetherbyusingadoptionsofthewell-knowniterativeclosestpointalgorithm.Usingaveryconstrainedcapturingsetup,Choietal.[61]performedastaticstitchingwithoutanycorrespondencemeasurement.ThesecondapproachbyLiuetal.[62],whichisalsousedbyAlkhathamietal.[31]onamobiledevice,extractsScaleInvariantFeatureTransformation(SIFT)featuresfrompreprocessedimagesandsearchesforcorrespondencesbetweenthem.Finally,theimagesarestitchedalongaborderlineandpost-processed.Toreachtheaimoftouchless-to-touchimageinteroperability,Salumetal.[63]proposedfurtherenhancementoftouchlessimagedata.Atfirst,theauthorsaddeddifferentrandomlychosenellipsestotheoriginalimage.Secondly,acontourenhancementbyahorizontalandverticalfadingisaddedtotheimage.Additionally,severalworksshowedthatridgethinningandskeletonizingapproachesfromthetouch-baseddomainarealsoapplicabletotouchlessimagedatatoimprovethebiometricperformance[25,27,55].QualitycontrolIncomparisontotouch-basedfingerprintrecognitionsystems,touchlessschemescontainmorecriticalstepsduringacquisitionandprocessingwhichcouldreducethesystemperformance.Forthisreason,anelaboratedqualityassuranceisparticularlyessentialfortouchlesssamples.Severalworksshowedthatdirectapplicationoftouch-basedfingerprintqualityassessmentleadstoinaccurateresults[64–66].Incontrast,Priesnitzetal.[67]demonstratedthatthetouch-basedqualityassessmenttoolNFIQ2.0isalsoapplicablefortouchlesssamples.Theauthorsconcludedthatthepredictivepowerhighlydependsonanadequatepre-processing.Figure 5adepictsafingerimageexampleofhighqualityincomparisontothreefingerimagesoflowqualityduetoacquisitionissues.InFig. 5b,theROIcontainsahighlightcausedbyanoverpoweredflashlightwhichleadstoalowrigde-valleycontrastwhilethecontrastonthewholefingerisratherhigh.AwrongfocuspositionresultsinablurryROIfromwhichnodetailsareextractableasshowninFig. 5c.FromarollposerotatedsampledepictedinFig. 5d,featuresareextractablebutnotcomparablewithanunrotatedpresentation. Fig.5Fingerimagequality.Exampleimagesofahigh-qualityfingerimageandthreelow-qualityfingerimagescapturedbyaSamsungGalaxyS8FullsizeimageForthepurposeofqualityassessment,differentauthorssuggesteddividingthefingerprintareaintoblocks.Subsequently,acertainqualityassessmentalgorithmisappliedtoeachoftheblockstoeithermergetheresultsofeachblocktoonescoreortoconsideronlyareasaboveacertainthresholdforfeatureextraction[7,42,66,68].ParzialeandChen[7]proposedacoherence-basedqualitymeasurement.Thisapproachmeasuresstrengthofthedominantdirectioninalocalregion.Forthispurpose,theauthorsappliedanormalizedcoherenceestimationonlocalgradientsofthegraylevelintensity.Moreover,thecovariancematrixofthegradientvectorswasdenotedwhichrepresentstheclarityoftheridgelinestructure.Lietal.[42,65]introducedaqualityassessmentalgorithmforfingerimagesacquiredwithsmartphones.Theauthorsuseddifferentmetricsinthespatialandfrequencydomainwhichresultedinafeaturevector.ASupportVectorMachine(SVM)wastrainedtoseparatehigh-qualityblocksfromthosewithlowquality.Yangetal.[66]presentedanotherqualitycontrolschemeforsamplescapturedinunconstrainedenvironments.Theinputfingerprintwasnotpreviouslysegmentedorprocessed.Thealgorithmusedtheamplitude-frequencyandridgelineorientationintheFourierdomainasdistinguishingqualityfeature.Eachblockreceiveditsownqualityvalue,soonlyhigh-qualityblockswereconsideredforfeatureextraction.Theauthorsconcludedthattheproposedalgorithmworksaccuratelyonthemajorityoftestedsamplesbutalsoprovidedfingerimageswhereitfails.ThesameauthorsextendedtheirapproachbyusinganSVM[68].Lietal.[69]furtherextendedtheamountofemployedqualityfeaturesbyadditionallyusingalocalclarityscoreandfrequencydomainanalysis.Leeetal.[51]proposedaneffectiveearlystagequalityestimationmethod.Theschemeisbasedongradientdistributionwhichshowsthecharacteristicsoftherepeatablelinepatternsofthefingerprintandthereforeitsquality.Forafirststagequalityestimation,thisschemeshowedagoodperformancecomparedinrelationtoitscomputationaleffort.AnothercontributionbyNohetal.[16]proposedacomparablequalityassessmentandridgefrequencyestimationandbenchmarkeditsperformance.Labatietal.[64]comparedtheirimplementationofaneuralnetworkclassificationsystemwithak-Nearest-Neighbor(kNN)classifier,alinear/quadraticdiscriminantclassifier,andNFIQ1.0[70].TheauthorsusedaratherconstraineddatasetandwereabletoshowthattheirownapproachperformssignificantlybetterthantheNFIQ1.0algorithm.Alatterworkofthesameauthorsshowedthecomputationalperformanceofthesysteminapracticalapproach[71].Zaghettoetal.[45]treatedrotationaldeviationsonmosaickedfingerprintscapturedinamulti-viewenvironmentasameasureofquality.Afour-layeredneuralnetworkwasproposedwhichclassifiestheinputdatasetintorotatedorun-rotated.FeatureextractionThefeatureextractionfromtouchlesscapturedfingerprintsamplesisperformedsimilarlytotouch-basedscenarios.Severalworksshowedthatestablishedfeatureextractorscanbeappliedtotouchlessimagedata,asshowninFig. 6.Whenusingtouch-basedalgorithms,itisimportanttonoticethatanextractorwhichperformsconsiderablygoodontouchlessandtouch-basedsamplesdoesnotnecessarilyleadtoaninteroperabilitybetweenthem.Touchlessdevelopmentsrangefromsimpletexturefeatureextractionwithout-of-the-boxalgorithmstodedicatedfingerprintfeatureextractors. Fig.6Featureextraction.Minutiaepointsextractedfromthetouch-basedfingerprint(a)andatouchlessfingerprint(b).ThefeatureextractionwasperformedwithFingerNet[72].Pleasenotethatduetothedifferentcapturingprocess,thetouchlessfingerprintimageismirroredFullsizeimageSomeworksinthetouchlessdomainusedthewell-establishedVerifingerSDKtoevaluatetheperformanceoftheirprocessingpipeline[37,73]orbenchmarkedtheirapproachesagainstit.Moreover,manyworksusedtheNISTstandardizedMINDTCT[74]algorithmforfeatureextractoronprocessedimages[18,24,41,63].Similarly,Yangetal.[66]usedthisfeatureextractorforqualityestimation.ItshouldbenotedthatVerifingerrequiresafingerprintscaledto500DPIinordertoworkproperly.ADPInormalizationasdescribedinSection7.4isusuallynotperformedbutcouldinfluencetheamountoffeaturesextracted.Hanetal.[73]investigatedthecompatibilityofphotographedfingerimageswiththeVerifingerfeatureextractor.TheauthorsshowedthatitispossibletoextractfeatureswithsomemanualpreprocessinginformofaROIextraction.ItshouldbenotedthatVerifingerdoesperformadditionalinternalpreprocessingwhichimprovestheoverallaccuracy.Sisodiaetal.[55]presentedasimplefeatureextractiontechniqueusingkerneloperationswhichrepresentcommonminutiaecharacteristics.TheworkproposedofRavietal.[27]describedanextractionandclassificationofminutiaecomparableto[55]usingthecountingnumberalgorithm.Onthepreprocessedbinaryimage,itcountstheamountofwhitepixelaroundthecenterpointandestimatesthecorrespondingminutiatype.AnotherworkbyWangetal.[75]appliedaslidingwindowonnormalizedimages.ItusedlocalgradientcodingsandLBPforfeatureextraction.Theauthorsanalyzeddifferentblocksizestoextractthetexturefeatures.Similarly,generalpurposetexturedescriptorshavebeenemployedin[76].Hiewetal.[77]transferredanapproachbasedonablock-wiseGabor-filterfromthetouch-baseddomaintotouchlessdata.Here,themagnitudewasconvertedtoascalarnumberwhichrepresentsthefeaturepoint.Inaddition,aPCAwasperformedtocompressthefeaturevectorandaprojectioninitsnormalizedEigenspaceisappliedtoeachGaborfeaturevector.Ramachandraetal.[18]usedSpectralMinutiaeRepresentation(SMR)onminutiaeextractedwithMINDTCTtoachieveafixedlengthfeaturevector.WithScatNet,Sankaranetal.[32]andMalhotraetal.[9]proposedanovelfeatureextractor.Group-invariantscatteringnetworks[78]refertoafilterbankofwaveletsthatproducearepresentationwhichwasshowntobestabletolocalaffinetransformations.Theauthorsextendedtheapproachwithanadditionalwavelet-modulustransformationforhighfrequencycomponents.Alow-passfilter-basedconvolutionconcatenatedthewaveletresponsesofanarbitrarynumberoffilterswhichleadtomorediscriminativefeatures.TheauthorscomparedtheirScatNetapproachtoaminutia-basedbaselineusingVeriFingerSDK[79]andMinutiaeCylinderCode(MCC)[80]forfeatureextractionandperformedslightlybetterthanthem.Yinetal.[81]proposedadistortion-freefeaturerepresentationusingtheridgecountitselfasfeature.Additionally,tosingleminutiae,pairsofminutiaewerealsoconsideredasfeature.Theauthorsusedageneticalgorithmtosolvethecombinatorialoptimizationproblem.Toimproveeffectivenessandaccuracy,aminutia-pairexpandingalgorithmwassuggested.Toperformcomparisonsonthesefeaturevectors,asimilaritymetricwasdefined.Ontwobenchmarkdatabases,theauthorswereabletoperformbetterthantheestablishedtouch-basedfeatureextractors.Itshouldbecriticallynotedthatintheirtestsetupthealgorithmhadahighoverallruntime.KumarandZhou[26]suggestedafeatureextractionbasedonlevel-0features,suchaslocaltexturepatterns.Theevaluationincludedvariouscombinationsofapproaches,e.g.,LocalizedRadonTransformation(LRT),andrevealedremarkablygoodperformance.Inamorerecentwork,VyasandKumar[82]suggestedanimprovedschemeusingminutiaecomparison.Genoveseetal.[23]proposedacombinationofimageprocessingalgorithmsandmachinelearningforextractinglevel-3features(sweatpores).TheauthorsextractedthegreenchannelfromanRGBimageandapplieddifferentgammatransformationsonit.Asimpleimageprocessingfollowedbyanextractionofconnectedcomponentsidentifiedcandidatesforsweatpores.ACNNdistinguishedwhetheracandidatepointisasweatporeornot.Buildinguponthiswork,Labatietal.[83]presentedacomparativestudyonlevel-3featureextraction.TwoCNNsweretrainedtodetectsweatporesonpreprocessedtouchless,touch-based,andlatentfingerprints.ThefirstCNNdeterminedpossiblesweatporesintheimageswhereasthesecondonedetectedfalselyselectedpores.Comparedtothetouch-basedresults,thetouchlessrecognitionperformanceturnedouttobeinferiorwhichwascausedbyvariableilluminationsituationsandporereflection.ComparisonInthefinalcomparisonstage,touchlessandtouch-basedfingerprintrecognitionsystemsoperateinasimilarway.Figure 7showsacomparisonofasinglefingerprintcapturedfromatouchlessandatouch-basedcapturingdevice.Similartothefeatureextractionstage,manyworksappliedcomparisonmethodsofthetouch-baseddomain,e.g.,theNISTbozorth3[74]comparator[41,63,84,85].TheNISTalsoevaluatedtheimpactoffingerprintsamplescapturedbytouchlessdevicesondifferentfingerprintrecognitionalgorithms[86]. Fig.7Minutiaecomparison.Manualcomparisonofminutiaeofatouch-basedfingerprintwithamirroredtouchlessfingerprintFullsizeimageLindosoetal.[87]introducedthefirstcomparatordedicatedtotouchlessfingerprintrecognitionin2007.Theauthorsproposedazeromeannormalizedcrosscorrelationapproach.Thismethodwasdirectlyappliedtothegraylevelsoftheinputimage.Inthefirststep,acoarsealignmentestimatedthewaytheimageswereshiftedandrotatedtofittothetemplate.Inthesecondstep,fingerprintregionswereselectedbasedonqualityandcomparedtoeachotherbasedonthegraylevelinafinalstep.Steinetal.[29]suggestedasimplecomparisonofallminutiaetoeachotherbasedontheModifiedHausdorffDistance(MHD)andorientation.KumarandZhou[26]comparedlevel-0featuresbyusinganormalizedHammingdistanceforanimagetexturecomparison.Theauthorsconcludedthatlocalizedfingerprintsub-regionsaremorerobusttorotationsandpartialdistortions.Labatietal.[88]presentedanapproachusingneuralnetworkstodetectapairofmatedminutiaebetweentwosamples.Alistoflocalfeaturesaroundanyminutiaeofthecorrespondingsamplewasestablished.Thisinformationwasincorporatedduringthetrainingoftheneuralnetwork.Itthendecidedifthecandidateswerereferringtothesameminutiaornot.Also,theworkincludesanalysesoncomparingmorethanonefingerprintview.Sankaranetal.[32]andMalhotraetal.[9]suggestedcombinationsofconventionalandmachinelearningtechniques.Atfirst,theconventionalalgorithmcomputedtheL1-distancebetweeneachtwoScatNetfeaturesresultinginacomparisonscore.Secondly,theapproachreliedonasupervisedbinaryclassifierwhichlearnedwhetheranimagepairisamatchornot.Buildingupontheirworkin[9],Malhortaetal.[89]showedthattheiralgorithmcanbeadaptedtoalsoworkonhighlyunconstraineddata.LinandKumar[90]proposedacomparisonframeworkbasedonamulti-SiameseCNNfortouchlesstotouch-basedfingerprintcomparison.Threesub-CNNsweretrainedonfingerprintminutiae,respectiveridgemaps,andspecificregionsofridgemaps.Theauthorsgenerateddeepfingerprintrepresentationswhichwereconcatenated.Thisapproachappearedtobemorerobustforcross-domaincomparisons.TheywereabletooutperformotherCNN-basedapproaches.AlaterworkbyTanandKumar[91]especiallyfocusedonposeinvariantfeaturematching.Toexploitthepropertiesoftheirintroducedfeaturesoptimally,Yinetal.[81]definedacomparisonmetricusinganumberofcorrespondingminutiaeandtheglobaltopologicalsimilarity.IssuesandchallengesInthepastyears,manyworksonthetopicoftouchlessfingerprintrecognitionhavebeenpublished.Nevertheless,therearestillsomeunsolvedissues.Thefollowingsubsectionssetoutthemostrelevantchallengesrelatedtothetouchlessrecognitionprocessandprovidestartingpointsforfurtherresearch.BiometricperformanceThemostimportantmeasurementcriterionforanybiometricsystemistherecognitionperformance.Table 4highlightsoutstandingtouchlessfingerprintrecognitionworkflowswiththeirachievedrecognitionperformance.Sofar,touchless2Dfingerprintschemesyieldaninferiorrecognitionaccuracycomparedtotouch-basedones.Practicalperformanceratesareonlyachievedbymoresophisticatedtouchlessapproaches,e.g.,basedon3Dfingerprintscapturedbysystemswhichutilizespecialacquisitiondevicesandcomprehensivepreprocessing[92].Uptonow,mobileapproachesusingacommoditydevicearenotabletoachievecompetitiveresults. Table4OverviewonselectedrecognitionworkflowswithbiometricperformanceFullsizetableAlongthetouchlessfingerprintrecognitionpipeline,differentstagesshouldbeconsideredtoachieveagoodbiometricperformance: Acquisition:Ahomogeneouslyilluminated,noise-freefingerimageshouldbeacquired.High-qualitycameraequipmentandapredictableilluminationareagoodpreconditionforaproperfingerimage. Preprocessing:Anaccuratelysegmentedandrotatedfingerprintimagesyieldmeaningfulcomparisonscores.Atthispoint,userinstructionsorafingerprintguidanceduringthecapturingprocesscanhelptoincreaseaccuracy. Qualityassessment:Adedicatedqualityassessmentwhichisintegratedinthepreprocessingpipelineiscrucialtoconsideronlysamplesofhighquality. Featureextractionandcomparison:Aspecifictouchlessfeatureextractionwhichisadaptedtotheconsidereddatasetrevealsresultscomparabletotouch-basedschemes. Also,itcanbeobservedthatsomeaspectsofthisresearchareahavebeenextensivelyresearched,whileothersdeservemoreattention.Forexample,severalwell-functioningsegmentationalgorithmshavebeenproposedwhereasonlylittleresearchhasbeenconductedondedicatedtouchlessfeatureextraction.EnvironmentalinfluencesTouchlessfingerprintcapturingandprocessinghastodealwithdifferentenvironmentalinfluences.Environmentalinfluencesorcomparisonbetweendifferentsensortypesmaylowertheperformance,asdiscussedinthefollowingsubsections.AccordingtoMalhotraetal.[9],challengingenvironmentalsituationare: Uncontrolledbackground Varyingillumination Fingerposition Impuritiesonthefingersurface Furthertechnicalchallengescanbesummarizedas: Varyingcamerasetup(especiallyonsmartphones) Noisyfingerprintimpressionduetolowcontrast Especiallyonmobiledevices,environmentalinfluenceshaveahighimpactonthebiometricrecognitionaccuracyasshowcasedbyMalhotraetal.[9].Fingerprintdetectionandsegmentationalgorithmshavetoberobustagainstahugevarietyofenvironmentalconditionsrangingfromverydarkenvironmentstooneswithbrightsunlight.Especiallycolor-basedsegmentationrevealsdeficitsonsceneswithabackgroundwhichcontainscolorsimilartoskincolor.Developersworkingonmobilesetupsshouldbeawareofthefactthatanacquisitionineveryenvironmentalsituationishardlyfeasible.Preprocessingandqualityassurancealgorithmsshouldbeabletoassessthesituationaspreciselyaspossibleandtodecidewhetherafingerprintcapturingisfeasible.Anappropriateuserfeedbackisexpectedtobehelpfulinsuchcases.Inprototypicalhardwaresetups,environmentalinfluencesplayaminorrole.Mostdeviceshaveahoodandhomogeneousbackgroundwhichensuresapredictableilluminationsituation,whereasothersrequirealaboratoryenvironmenttoworkproperly[16].SetupsdesignedfortheusageunderdifferentenvironmentalinfluencecouldalsobenefitfromtheuseofdepthinformationonanimagelikesuggestedbyJonitzandJivet[58].Theadditionaldepthinformationhelpsalgorithmstosegmentthefinerandgivesahintonthedistancebetweenfingerandsensor.UsabilityandacceptabilityOneofthemainadvantagesoftouchlessfingerprintacquisitionisseeninahigherusabilitycomparedtotouch-basedschemes.Touch-basedfingerprintcapturingsuffersfromhygienicissuesincasevariousparticipantsaretouchingthesensorsurface.Touch-basedschemesalsorequireacertainorientationandpressureofthefingerandgenerallyneedmoretimeforthecapturingprocess.AsdiscussedinSection2,touchlesscapturingdevicesshowdifferentlevelsofusability.Ingeneral,ahigherusabilitycanbeachievedby: 1 Sensor-to-fingerdistance:Afreelychosendistancebetweenthefingerandsensorduringthepresentationofthefingerisdesirable. 2 Poseangle:Anunconstrainedorientationduringthepresentationofthefingerleadstoamoreconvenientsystem. 3 Fourprintcapturing:Mosttouchlessdevicescandirectlycaptureuptofourfingersinoneacquisitionprocess.Preprocessingisthenabletoaccuratelyseparatethefingerprintareasintofingerprintimages. 4 Integratedqualityassessment:Anintegratedqualitymeasureensuresthatthecapturingprocessisfinishedassoonasonehigh-qualitytemplateofoneormorefingeriscaptured. 5 Fastcapturingprocess:Thetimeneededtopresentthefingersaccuratelyshouldbeasshortaspossible.Processingstepsshouldbeappliedsubsequenttoacquisitionwhereveritisfeasible. 6 Easy-to-understanduserfeedback:Anintegrateduserfeedbackhelpstopresentthefingerssmoothly. Thepoints1–4addressanunconstrainedacquisitionprocesswhichishighlydesirableforenhancedusability.Nevertheless,amoreunconstrainedcapturingalsorequiresmorerobustfingerdetectionalgorithmsandespeciallyanelaboratedqualityassessmenttoavoidthecapturingoflow-qualitysamples.Theseusabilitygoalscanonlybeachievedwithanlargeamountofprocessingpower.Today,nomobilecapturingsetupsatisfiesalloftheserequirements.Themajorityofcommoditydevicesforcapturingfocusonaratherunconstrainedcapturing(e.g.,[33])whereasprototypicalhardwaresetupsfocusmoreonrecognitionaccuracy[16].Inacomprehensivestudy,Furmanetal.[93]evaluatedtheusabilityofthreestationarytouchlessrecognitionproducts.Theauthorscametotheconclusionthattouchlesscapturingrequiresadedicatedinstruction.Touchless-to-touch-basedsensorinteroperabilityInteroperabilitybetweentouch-basedandtouchlesssensorsisadesirableobjectiveinmanycases,e.g.,toavoidre-enrolmentofsubjectsalreadyregisteredwiththesystemincaseofsensorexchangeortoenablecross-matchingbetweenfingerprintdatabasescapturedthroughtouchlessandtouch-basedsensors.Afundamentaldifferencebetweentouch-basedandtouchlessfingerprintsisthattouchlessfingerprintsaremirroredalongtheverticalaxis.Themajorityoftouchlesssensorsalsocapturecolorfingerimageswhereastouch-basedsensorscapturegrayscalefingerprints.Further,touchlessfingerprintscontainnodeformationsduetopressingthefingerontoasurface.Somedifferences,e.g.,mirroring,color-to-grayscaleconversionorinvertedback-andforeground,canbeimplementedinastraight-forwardmannerwithoutalossofaccuracy.Otherdifferencesrequireelaboratedapproximationapproaches,e.g.,theaspectratioordeformationestimation[94].Anaccurateandrobustschemeforcorrectingdeformationsontouchless2Dfingerprintimageshasnotyetbeenestablished.OneimportantfactorwhichmaycausebiometricperformancedropsininteroperabilityscenariosistheDPIalignmentfortouchlessdata.Fortouch-basedsensorsthemeasureofspatialdotdensityisanimportantmetricforacquisitiondevicestoalignthedatasamplestoacertainsizeandresolution.ISO/IECcompliantfingerprintsneedtoexhibit500DPIwhichnowadaysisaminimumrequirementforcommercialproducts[95].TouchlessdevicessuchasdigitalcamerasfeaturenoDPIvaluebecausetheacquiredimageisnotboundtoaphysicalscale.Nonetheless,itismandatorytonormalizetouchlessfingerprintstothesamesizeandresolutioninordertoachieveanaccurateperformance.Fingerprintimagescanbenormalizedbycroppingtheimageareaandrescalingittoacertainheightandwidth.Byknowingthesensorsresolutionandfocallengthandapproximatingthedistancebetweenfingerandsensorviatheautofocusandthefingers’widththeDPIofthefingerareacanbeapproximatedtoanalmostconstantvalue[33,61].Wildetal.[96]proposedacomparativetestoftheirresolutionestimationschemeondifferentsmartphones.Theauthorswereabletoachieveaccuratecomparisonscoresinaninteroperabilityscenario.Anotherimportantissueistheridgefrequencyestimationontouchlessdata.Theridgefrequencyofafingerprintreferstotheamountofridgeswhicharepresentwithinawindowofdefinedsize.Duetothetouchlessacquisition,thereisnodeformationresultingfrompressingthefingerontothesensorsurface.Considering2Dfingerprintimagesthismeansthatthefrequencyofridgesisincreasingtowardsthebordersincontrasttotouch-basedfingerprintswhereitstaysalmoststable.Moreover,blurredborderareasflattenthepeakswhichhamperscorrectfeaturedetection.Thinplatesplinesareasuitabletooltocorrectthesedeformationsingeneralwhichalsohasapositiveeffectontheridgefrequencyandinteroperability[16,48].Inafirstapproach,thealgorithmofNohetal.[16]searchedforcorrespondingpointsintouchlessandtouch-basedsamplesandminimizesanenergyfunction.Thisapproachshowedaccurateresultsbutishardlypracticallyimplementablebecauseonetouchlessandonetouch-basedsampleisneeded.Linetal.[48]wentonestepfurtherandformulatedadeformationcorrectionmodelbasedonrobustthinplatesplines.Differentmodelsweretrainedtomeettheindividualfingershape.Duringthecomparisondifferentdeformationcorrectionmodelswereautomaticallyselected.AcomparablemethodwasalsosuggestedbyDaboueietal.[97].TheNISTalsoconductedacomprehensivestudyoninteroperabilityissuesinapplicationscenariosweretouchlessandtouch-basedfingerprintsarecompared[98].PresentationattackdetectionReliablePresentationAttackDetection(PAD),i.e.,anti-spoofing,modulesarevitaltoenhancethesecurityoffingerprintrecognitionsystems.PADrepresentsawell-studiedfieldofresearchfortouch-basedfingerprintrecognitionsystems[99].Specializedhardware-basedskindetectionmethodswhicharereportedtoreliablydetectdiversePresentationAttackInstruments(PAI)species,e.g.,gummyfingers,arealreadyintegratedinmanycommercialtouch-basedfingerprintcapturingdevices.Incontrast,inatouchlessfingerprintrecognitionsystem,PADturnsouttobemorechallenging.Upuntilnow,onlyafewapproachestoPADintouchlessfingerprintacquisitionhavebeenproposed.Moonetal.[100]proposedaPADmethodbasedonwaveletanalysisofthefingertipsurfacetexture.Wangetal.[15]presentedaPADalgorithmwhichexploitsthedifferencesbetweenbonafidepresentationsandattackpresentationsinband-selectiveFourierspectra.Inaddition,reflectiondetectionwasimplementedtodetectfakefingermaterials.Avideo-basedPADmethodbasedonthedetectionofsweatporeswaspresentedbyParzialeandChen[7].TheideaofPADfortouchlessfingerprintacquisitionusingtexturedescriptorsinconjunctionwithneuralnetwork-basedclassifierswasproposedbyAlkhathamietal.[31].Moreover,adetectionoffingerveinscanbeemployedforPADinatouchlessfingerprintrecognitionsystem.AnapproachforPADwithasetupbasedonsmartphonesispresentedbySteinetal.[30].Theyusedavideo-basedacquisitionandshowthatitispossibletodetectpresentationattacksbyanalyzingdifferentvideoframes.AfurtherworkbyOvergaardetal.[101]triedtoexploitEulerianVideoMagnification(EVM)forlivenessdetection.Themethodemphasizedtheheartbeat-relatedcolorvariationsofgenuinefingers.However,theauthorsraisedseveralconcernsthatthisapproachmightnotbeputintopractice.Tanejaetal.[102]createdalargepubliclyavailablespoofedfingerphotodatabase.Thedatabasecontainsprint-outattacks,photoattacks,andnon-spoofedfingerimagescapturedwithtwodifferentsmartphones.BiometrictemplateprotectionDuetothestrongandpermanentlinkbetweenindividualsandtheirfingerprints,exposureofenrolledfingerprinttemplatestoadversariescanseriouslycompromisebiometricsystemsecurityanduserprivacy,e.g.,stolenfingerprintscouldbeusedtocreateartifactsinordertolaunchpresentationattacks.Numeroustechniqueshavebeenproposedforfingerprint-basedbiometrictemplateprotectionoverthelast20years[103,104].Inaddition,theISO/IECstandardfortheprotectionofbiometricinformation[105]providesguidanceforprotectionunderrequirementsofconfidentiality,integrity,andrenewability/revocabilityduringstorageandtransferandforsecureandprivacy-compliantmanagementandprocessingofbiometricinformation.Whileoriginallydesignedandevaluatedontouch-basedfingerprintdatabases,conceptsforbiometriccryptosystems,e.g.,thefuzzyvaultscheme[106,107]orthefuzzycommitmentscheme[108,109],andcancelablebiometrics,e.g.,Cartesian,radialorfunctionaltransformations[110,111],couldbeadaptedtotouchlessfingerprints,too.Dependingontheemployedscheme,featuretypetransformationsoffingerprinttemplatesmightberequired[112].Duetothisreason,almostnoresearchhasbeenconductedtodesignparticulartemplateprotectionschemesfortouchlessfingerprints.Mostnotably,Hiewetal.[77]proposedtheuseofmultiplerandomprojectionstoachieveacancelabletouchlessfingerprintrecognitionsystem.Similarly,Zannouetal.[113]suggestedaschemeforrevocabletouchlessfingerprinttemplateextraction.Laietal.[114]presentedanalgorithmwhichdirectlyencryptsfingerprintimagesusinganovelmemristivechaoticsystem.Malhotraetal.[115]addressedtheissueoffingerprinttemplateprotectioninselfieimagesonsocialmediaplatforms.Multi-biometricsMulti-biometricsystemshavebeenfoundtosignificantlyimprovetheaccuracyandreliabilityofbiometricsystems[116].Withthepossibilityofaslaphandacquisition,thefusionofbiometricinformationobtainedfromfourfingerscanbeemployedtoimprovebiometricperformance,especiallyinunconstrainedenvironments.Debetal.[34]demonstratedthepotentialoffusinginformationoffourfingersacquiredthroughtwoslaphandacquisitiondevices.Nohetal.[117]proposedascore-levelfusionofthreefingersacquiredbyatouchlesssensortoachievehigherrecognitionaccuracy.Carneyetal.[33]performedascore-levelfusionoftwo,four,andeightfingers.Theywereabletoachievesignificantperformancegainsduetothefusion.Moreover,biometricinformationobtainedfromtouchlessfingerprintscouldbefusedwithdifferentbiometriccharacteristics.Improvementinbiometricperformanceasaresultofbiometricfusionshouldbeweighedagainsttheassociatedoverheadinvolved,suchasadditionalsensingcost,i.e.,itispreferredtocombinebiometriccharacteristicsthatcanbeacquiredinasinglepresentation[118].Mil’shteinetal.[14]andRamachandraetal.[18]suggestedafusionoffingerveinpatternswithtouchlessfingerprints.ResearchresourcesDatabasescomprisingtouchlessfingerprintimagedataarevitalforthedevelopmentofimprovedprocessingmodules.AnoverviewofdatabasesavailableforresearchpurposesandtheirpropertiesisgiveninTable 5. Table5OverviewofpubliclyavailabletouchlessfingerprintdatabasesFullsizetableTheHongKongPolytechnicUniversityestablishedseveraldatabasesfordifferentproposals.Sofar,themostcomprehensivetouchless-to-touchfingerprintdatabasehasbeenestablishedbyKumar[120].Itconsistsof1800touchless2Dfingerimagesandthecorrespondingtouch-basedfingerprintsacquiredfrom300subjects.Amultimodaldatabase[121]features62642Dfingerimagesincludingcorrespondingveinimagesof156subjectsareprovidedwith6samplesofindexandmiddlefingersastextureandveinimageforeachsubject.Anotherdatabasecontaininglow-resolutionfingersurfaceimagesacquiredbyalow-costwebcamwasestablishedin[122].Thedatabasecontains1466imagesfrom156subjectscapturedintwosessions.TheIIITDSmartPhoneFingerphotoDatabasev1(ISPFDv1)[32]isasmartphonefingerphotodatabasewhichconsistsof4096fingerphotoimagesfrom128subjects.Thedatabaseisacquiredusingasmartphonecamerawithvaryingbackgroundandillumination.Persubject8,imagesofboth,therightindexandmiddlefinger,aretaken.Theilluminationiscategorizedinindoorandoutdoorwhereasthebackgroundisseparatedintoawhiteoneandabusyone.Everycategorycontainstwofingersintwolightningandbackgroundsituations.Insummary,4096imagesweretakenandadditionallyacquiredwithatouch-baseddevicetoestimatethecross-sensorcomparisonperformance.Afollow-updatabaseISPFDv2[89]wascapturedusingtwosmartphonesandonetouch-baseddevice.Itincludesmorethan17,000touchlessand2432touch-basedsamplesof304fingers.Afurtherextensionbypresentationattacksisproposedbythesameinstitution[102].Theauthorscaptured128presentationattacksusingopticaldevicesandprinters.TheSocial-MediaPostedFinger-selfie(SMPF)database[102]provides1000imagesdownloadedfromsocialmediaplatformswhichcontainfingers.Thisdatabasecouldbeusedforresearchontemplateprotectionschemes.Chopraetal.[123]collectedanothersmartphone-baseddatabase.TheUNconstrainedFIngerphoTo(UNFIT)databasecontains3450samplesof115subjects,capturedusingmultiplesmartphoneswithdifferentresolutions.Thesamplesarecapturedconsideringdifferentchallenges,suchasbackground,illumination,miss-focusingandmulti-fingerpresentations.Thisdatabaseiswell-suitedforresearchonfingerdetectionandqualityaspectsbutinappropriateforbiometricperformancetesting.IITBombay,TouchlessandTouch-BasedFingerprintDatabase[35]consistsof800touchlessand800touch-basedfingerprintimagesof200subjects.Thetouchlesssamplesarecapturedusingasmartphonewiththedevelopedandroidappandarecroppedtoanimagesizeof170×260.Thedatabasealsoconsistsof800touch-basedfingerprintsofthesame200subjectswithanimagesize260×330.Itaimstohelpresearchersintheirendeavorsincomparingtheperformanceoftouchlessandtouch-basedfingerprintbiometricsystems.ThefirstsmartphonespoofingattackdatabasebyTanejaetal.[102]contains4096bonafidefingerimagesand8182spoofingattacks.ThebonafideimagesaretakenfromtheISPFDv1database.Fromthedataset,theauthorscreated2048printattacks(printoutswhichwereagainphotographed)and6144photoattacks.ThephotoattacksaretakenfromthescreensofaniPad,asmartphone,andalaptop.TheauthorsusedthesamedevicesasintheISPFDv1database.Thesemi-publicFootnote1cross-sensorGUC100database[124]containsfivetouch-basedandonetouchlesssensor(TSTBird3).Duringthedatabaseestablishment100subjectspresentedtheir10fingerstoall6devices.Thiswasrepeated12,toobtainnaturalvariance.Allinallapproximately72,000imageswerecollected.ConclusionsInthiswork,thestate-of-the-artintheconstantlyevolvingfieldoftouchlessfingerprintrecognitionissummarizedanddiscussed.Thisresearchfieldfeaturesabroadspectrumofdifferentacquisitionsystemsfromhigh-endsetupstolow-costdevices.Subsequently,differentpreprocessingapproacheshavetobeappliedtotheacquiredimagedata.Itcanbeobservedthatageneralendeavorofsummarizedresearchistoachieveinteroperabilitybetweentouchlessandtouch-basedfingerprintrecognitionsystems.Ingeneral,touchlessschemesrevealimprovedusabilityandhighuseracceptancewhereasbiometricperformanceremainsaschallenge,especiallyonmobileof-the-shelfdevices.Conceptsforfurtherresearchtopicsrelatedtotouchlessfingerprintrecognition,e.g.,PADorbiometrictemplateprotection,havealreadybeenpresentedintheliterature.Buildingupontheseconcepts,firststationaryandmobilecommercialtouchlessfingerprintrecognitionsystemshavebeenintroduced.However,moreworkisyettobedoneinordertoachieverobust,interoperable,secure,privacypreserving,anduser-friendlysystems. Availabilityofdataandmaterials Datasharingnotapplicabletothisarticleasnodatasetsweregeneratedoranalyzedduringthecurrentstudy. NotesThedatabaseisnotpubliclyavailablebutresearcherscansendintheiralgorithmstotestthemonthedatabase.ReferencesA.K.Jain,P.Flynn,A.A.Ross,HandbookofBiometrics,1stedn(Springer,Boston,2008).Book  GoogleScholar  D.Maltoni,D.Maio,A.Jain,S.Prabhakar,HandbookofFingerprintRecognition.Chap.Fingerprintanalysisandrepresentation,2ndedn(Springer,London,2009).MATH  Book  GoogleScholar  A.K.Jain,K.Nandakumar,A.Ross,50yearsofbiometricresearch:Accomplishments,challenges,andopportunities.PatternRecogn.Lett.79:,80–105(2016).Article  GoogleScholar  R.D.Labati,F.Scotti,inEncyclopediaofCryptographyandSecurity.Fingerprint(SpringerUSBoston,2011),pp.460–465.Chapter  GoogleScholar  R.D.Labati,A.Genovese,V.Piuri,F.Scotti,Touchlessfingerprintbiometrics:Asurveyon2dand3dtechnologies.J.InternetTechnol.15(3),328(2014). GoogleScholar  Y.Song,C.Lee,J.Kim,inInternationalSymposiumonIntelligentSignalProcessingandCommunicationSystems.Anewschemefortouchlessfingerprintrecognitionsystem(IEEENewYork,2004),pp.524–527. GoogleScholar  G.Parziale,Y.Chen,inHandbookofRemoteBiometrics.Advancedtechnologiesfortouchlessfingerprintrecognition(SpringerLondon,2009),pp.83–109.Chapter  GoogleScholar  M.S.Khalil,F.K.Wan,inInternationalConferenceonWaveletAnalysisandPatternRecognition(ICWAPR).Areviewoffingerprintpre-processingusingamobilephone(IEEENewYork,2012),pp.152–157. GoogleScholar  A.Malhotra,A.Sankaran,A.Mittal,M.Vatsa,R.Singh,inHumanRecognitioninUnconstrainedEnvironments.Chapter6-fingerphotoauthenticationusingsmartphonecameracapturedundervaryingenvironmentalconditions(AcademicPressNewYork,2017),pp.119–144.Chapter  GoogleScholar  S.Mil’shtein,A.Pillai,inIEEEInternationalSymposiumonTechnologiesforHomelandSecurity(HST).Perspectivesandlimitationsoftouchlessfingerprints(IEEENewYork,2017),pp.1–6. GoogleScholar  R.DonidaLabati,A.Genovese,V.Piuri,F.Scotti,inSelfieBiometrics:AdvancesandChallengesAdvancesinComputerVisionandPatternRecognition.ASchemeforFingerphotoRecognitioninSmartphones(SpringerCham,2019),pp.49–66.Chapter  GoogleScholar  ISO/IECJTC1SC37Biometrics,ISO/IEC2382-37:2017InformationTechnology-Vocabulary-Part37:Biometrics(InternationalOrganizationforStandardization,Geneva,2017). GoogleScholar  J.Libert,J.Grantham,B.Bandini,S.Wood,M.Garris,K.Ko,F.Byers,C.Watson,Guidanceforevaluatingcontactlessfingerprintacquisitiondevices.NISTSpec.Publ.500:,305(2018). GoogleScholar  S.Mil’shtein,M.Baier,C.Granz,P.Bustos,inIEEEConferenceonTechnologiesforHomelandSecurity.Mobilesystemforfingerprintingandmappingofblood-vesselsacrossafinger(IEEENewYork,2009),pp.30–34. GoogleScholar  L.Wang,R.H.A.El-Maksoud,J.M.Sasian,W.P.Kuhn,K.Gee,V.S.Valencia,inNovelOpticalSystemsDesignandOptimizationXII,7429.Anovelcontactlessaliveness-testing(cat)fingerprintsensor(SPIEBellingham,2009),pp.333–343. GoogleScholar  D.Noh,H.Choi,J.Kim,Touchlesssensorcapturingfivefingerprintimagesbyonerotatingcamera.Opt.Eng.50(11),113202(2011).Article  GoogleScholar  C.Tsai,P.Wang,J.Yeh,inNovelOpticalSystemsDesignandOptimizationXVII,9193.Compacttouchlessfingerprintreaderbasedondigitalvariable-focusliquidlens(SPIEBellingham,2014),pp.173–178. GoogleScholar  R.Raghavendra,K.B.Raja,J.Surbiryala,C.Busch,Alow-costmultimodalbiometricsensortocapturefingerveinandfingerprint.Int.JtConf.Biom.IEEE,1–7(2014).A.Weissenfeld,B.Strobl,F.Daubner,in2018Design,AutomationTestinEuropeConferenceExhibition(DATE).Contactlessfingerandfacecapturingonasecurehandheldembeddeddevice(IEEENewYork,2018),pp.1321–1326.Chapter  GoogleScholar  J.Palma,C.Liessner,S.Mil’Shtein,Contactlessopticalscanningoffingerprintswith180∘view.Scan.28(6),301–304(2006).Article  GoogleScholar  C.Lee,S.Lee,J.Kim,S.-J.Kim,inAdvancesinBiometrics.Preprocessingofafingerprintimagecapturedwithamobilecamera(SpringerBerlin,2005),pp.348–355.Chapter  GoogleScholar  B.Y.Hiew,A.B.J.Teoh,Y.H.Pang,inInternationalConferenceonTelecommunicationsandMalaysiaInternationalConferenceonCommunications.Digitalcamerabasedfingerprintrecognition(IEEENewYork,2007),pp.676–681. GoogleScholar  A.Genovese,E.Munoz,V.Piuri,F.Scotti,G.Sforza,inIEEECongressonEvolutionaryComputation(CEC).Towardstouchlessporefingerprintbiometrics:aneuralapproach(IEEENewYork,2016),pp.4265–4272. GoogleScholar  V.Piuri,F.Scotti,inSecondInternationalConferenceonBiometrics:Theory,ApplicationsandSystems(BTAS).FingerprintBiometricsviaLow-costSensorsandWebcams(IEEENewYork,2008),pp.1–6. GoogleScholar  R.Mueller,R.Sanchez-Reillo,inFifthInternationalConferenceonIntelligentInformationHidingandMultimediaSignalProcessing(IIH-MSP).AnApproachtoBiometricIdentityManagementUsingLowCostEquipment(IEEENewYork,2009),pp.1096–1100. GoogleScholar  A.Kumar,Y.Zhou,inConferenceonComputerVisionandPatternRecognitionWorkshops(CVPRW).Contactlessfingerprintidentificationusinglevelzerofeatures(IEEENewYork,2011),pp.114–119. GoogleScholar  H.Ravi,S.K.Sivanath,inInternationalConferenceonTechnologiesforHomelandSecurity(HST).Anovelmethodfortouch-lessfingerprintauthentication(IEEENewYork,2013),pp.147–153. GoogleScholar  M.O.Derawi,B.Yang,C.Busch,inSecurityandPrivacyinMobileInformationandCommunicationSystems(ICST).Fingerprintrecognitionwithembeddedcamerasonmobilephones(SpringerBerlinHeidelbergBerlin,2012),pp.136–147. GoogleScholar  C.Stein,C.Nickel,C.Busch,inInternationalConferenceofBiometricsSpecialInterestGroup(BIOSIG).Fingerphotorecognitionwithsmartphonecameras(IEEENewYork,2012),pp.1–12. GoogleScholar  C.Stein,V.Bouatou,C.Busch,inInternationalConferenceoftheBiometricSpecialInterestGroup(BIOSIG).Video-basedfingerphotorecognitionwithanti-spoofingtechniqueswithsmartphonecameras(IEEENewYork,2013),pp.1–12. GoogleScholar  M.Alkhathami,F.Han,R.VanSchyndel,inProceedingsoftheInternationalConferenceonDistributedSmartCamerasInternationalConferenceonDistributedSmartCameras(ICDSC).AMosaicApproachtoTouchlessFingerprintImagewithMultipleViews(AssociationforComputingMachineryNewYork,2014),pp.22–1228. GoogleScholar  A.Sankaran,A.Malhotra,A.Mittal,M.Vatsa,R.Singh,in7thInternationalConferenceonBiometricsTheory,ApplicationsandSystems(BTAS).Onsmartphonecamerabasedfingerphotoauthentication(IEEENewYork,2015),pp.1–7. GoogleScholar  L.A.Carney,J.Kane,J.F.Mather,A.Othman,A.G.Simpson,A.Tavanai,R.A.Tyson,Y.Xue,in4thInternationalConferenceonBiomedicalandBioinformaticsEngineering(ICBBE).Amulti-fingertouchlessfingerprintingsystem:Mobilefingerphotoandlegacydatabaseinteroperability(AssociationforComputingMachineryNewYork,2017),pp.139–147. GoogleScholar  D.Deb,T.Chugh,J.Engelsma,K.Cao,N.Nain,J.Kendall,A.K.Jain,Matchingfingerphotostoslapfingerprintimages.arXivpreprintarXiv:1804.08122(2018).P.Birajadar,M.Haria,P.Kulkarni,S.Gupta,P.Joshi,B.Singh,V.Gadre,Towardssmartphone-basedtouchlessfingerprintrecognition.Sādhanā.44(7),161(2019).Article  GoogleScholar  B.Y.Hiew,A.B.J.Teoh,Y.H.Pang,inWorkshoponAutomaticIdentificationAdvancedTechnologies(AutoID).Touch-lessfingerprintrecognitionsystem(IEEENewYork,2007),pp.24–29. GoogleScholar  P.Birajadar,S.Gupta,P.Shirvalkar,V.Patidar,U.Sharma,A.Naik,V.Gadre,inInternationalConferenceonSignalandInformationProcessing(IConSIP).Touch-lessfingerphotofeatureextraction,analysisandmatchingusingmonogenicwavelets(IEEENewYork,2016),pp.1–6. GoogleScholar  S.Sagiroglu,M.Ulker,B.Arslan,inCongressonEvolutionaryComputation(CEC).MobileTouchlessFingerprintAcquisitionAndEnhancementSystem(IEEENewYork,NY,2020),pp.1–8. GoogleScholar  P.Wasnik,R.Raghavendra,M.Stokkenes,K.Raja,C.Busch,inInternationalConferenceoftheBiometricsSpecialInterestGroup(BIOSIG).ImprovedFingerphotoVerificationSystemUsingMulti-scaleSecondOrderLocalStructures(IEEENewYork,NY,2018),pp.1–5. GoogleScholar  A.Weissenfeld,A.Zoufal,C.Weiss,B.Strobl,G.F.Dominguez,inEuropeanIntelligenceandSecurityInformaticsConference(EISIC).TowardsMobileContactless4-FingerprintAuthenticationforBorderControl(IEEENewYork,2018),pp.73–76. GoogleScholar  R.Raghavendra,C.Busch,B.Yang,inSixthInternationalConferenceonBiometrics:Theory,ApplicationsandSystems(BTAS).Scaling-robustfingerprintverificationwithsmartphonecamerainreal-lifescenarios(IEEENewYork,2013),pp.1–8. GoogleScholar  G.Li,B.Yang,M.AastrupOlsen,C.Busch,inConferenceonComputerVisionandPatternRecognitionWorkshops(CVPRW).Qualityassessmentforfingerprintscollectedbysmartphonecameras(IEEENewYork,2013),pp.146–153. GoogleScholar  K.Tiwari,P.Gupta,inInternationalConferenceonBiometrics(ICB).Atouch-lessfingerphotorecognitionsystemformobilehand-helddevices(IEEENewYork,2015),pp.151–156. GoogleScholar  K.Wang,Y.Cao,X.Xing,inBiometricRecognition.Contrastresearchonfullfingerareaextractionmethodoftouchlessfingerprintimagesunderdifferentilluminants(SpringerInternationalPublishingCham,2017),pp.269–278.Chapter  GoogleScholar  C.Zaghetto,A.Zaghetto,F.d.B.Vidal,L.H.M.Aguiar,inInternationalConferenceonBiometrics(ICB).Touchlessmultiviewfingerprintqualityassessment:rotationalbad-positioningdetectionusingArtificialNeuralNetworks(IEEENewYork,2015),pp.394–399. GoogleScholar  K.Wang,H.Cui,Y.Cao,X.Xing,R.Zhang,inBiometricRecognition.Apreprocessingalgorithmfortouchlessfingerprintimages(SpringerInternationalPublishingCham,2016),pp.224–234.Chapter  GoogleScholar  R.D.Labati,A.Genovese,V.Piuri,F.Scotti,inInternationalConferenceonComputationalIntelligenceforMeasurementSystemsandApplications(CIMSA).Measurementoftheprincipalsingularpointincontactandcontactlessfingerprintimagesbyusingcomputationalintelligencetechniques(IEEENewYork,2010),pp.18–23. GoogleScholar  C.Lin,A.Kumar,MatchingContactlessandContact-basedConventionalFingerprintImagesforBiometricsIdentification.Trans.ImageProcess.27(4),2008–2021(2018).MathSciNet  MATH  Article  GoogleScholar  D.Lee,W.Jang,D.Park,S.-J.Kim,J.Kim,inFourthWorkshoponAutomaticIdentificationAdvancedTechnologies(AutoID).Areal-timeimageselectionalgorithm:fingerprintrecognitionusingmobiledeviceswithembeddedcamera(IEEENewYork,2005),pp.166–170. GoogleScholar  J.M.Tenenbaum,Accommodationincomputervision(TechRep.StanfordUnivCaDeptofComputerScience,1970).D.Lee,K.Choi,H.Choi,J.Kim,Recognizable-imageselectionforfingerprintrecognitionwithamobile-devicecamera.Trans.Syst.ManCybern.BCybern.38(1),233–243(2008).Article  GoogleScholar  C.Jonietz,E.Monari,H.Widak,C.Qu,in12thInternationalConferenceonAdvancedVideoandSignalBasedSurveillance(AVSS).Towardsmobileandtouchlessfingerprintverification(IEEENewYork,2015),pp.1–6. GoogleScholar  B.Y.Hiew,A.B.J.Teoh,D.C.L.Ngo,inInternationalConferenceonComputerGraphics,ImagingandVisualisation(CGIV).AutomaticDigitalCameraBasedFingerprintImagePreprocessing(IEEENewYork,2006),pp.182–189.Chapter  GoogleScholar  A.Cheddad,J.Condell,K.Curran,P.M.Kevitt,in16thInternationalConferenceonImageProcessing(ICIP).Anewcolourspaceforskintonedetection(IEEENewYork,2009),pp.497–500. GoogleScholar  D.S.Sisodia,T.Vandana,M.Choudhary,inInternationalConferenceonPower,Control,SignalsandInstrumentationEngineering(ICPCSI).Aconglomeratetechniqueforfingerprintrecognitionusingphonecameracapturedimages(IEEENewYork,2017),pp.2740–2746.Chapter  GoogleScholar  J.Priesnitz,C.Rathgeb,N.Buchmann,C.Busch,in26thInternationalConferenceonPatternRecognitionWorkshop(ICPRW).DeepLearning-basedSemanticSegmentationforTouchlessFingerprintRecognition(SpringerInternationalPublishingBasel,2021),pp.207–216. GoogleScholar  T.B.Fitzpatrick,TheValidityandPracticalityofSun-ReactiveSkinTypesIThroughVI.Arch.Dermatol.124(6),869–871(1988).Article  GoogleScholar  C.Jonietz,I.Jivet,inInternationalSymposiumonElectronicsandTelecommunications(ISETC).Touchlessfingerprintcapturingfromrgb-dimagesinmobiledevices(IEEENewYork,2018),pp.1–4. GoogleScholar  A.K.Jain,S.Prabhakar,L.Hong,Amultichannelapproachtofingerprintclassification.IEEETrans.PatternAnal.Mach.Intell.21(4),348–359(1999).Article  GoogleScholar  X.Liu,M.Pedersen,C.Charrier,F.A.Cheikh,P.Bours,in6thEuropeanWorkshoponVisualInformationProcessing(EUVIP).Animproved3-stepcontactlessfingerprintimageenhancementapproachforminutiaedetection(IEEENewYork,2016),pp.1–6. GoogleScholar  H.Choi,K.Choi,J.Kim,Mosaicingtouchlessandmirror-reflectedfingerprintimages.Trans.Inf.ForensicsSecur.5(1),52–61(2010).Article  GoogleScholar  F.Liu,D.Zhang,C.Song,G.Lu,Touchlessmultiviewfingerprintacquisitionandmosaicking.IEEETrans.Instrum.Meas.62(9),2492–2502(2013).Article  GoogleScholar  P.Salum,D.Sandoval,A.Zaghetto,B.Macchiavello,C.Zaghetto,inInternationalConferenceonImageProcessing(ICIP).Touchless-to-touchfingerprintsystemscompatibilitymethod(IEEENewYork,2017),pp.3550–3554. GoogleScholar  R.D.Labati,V.Piuri,F.Scotti,inInternationalJointConferenceonNeuralNetworks(IJCNN).Neural-basedqualitymeasurementoffingerprintimagesincontactlessbiometricsystems(IEEENewYork,2010),pp.1–8. GoogleScholar  G.Li,B.Yang,C.Busch,in201318thInternationalConferenceonDigitalSignalProcessing(DSP).AutocorrelationandDCTbasedqualitymetricsforfingerprintsamplesgeneratedbysmartphones(IEEENewYork,2013),pp.1–5. GoogleScholar  B.Yang,X.Li,C.Busch,inMultimediaonMobileDevices2012;andMultimediaContentAccess:AlgorithmsandSystemsVI,8304.Collectingfingerprintsforrecognitionusingmobilephonecameras(SPIEBellingham,2012),pp.182–189. GoogleScholar  J.Priesnitz,C.Rathgeb,N.Buchmann,C.Busch,inInternationalConferenceoftheBiometricsSpecialInterestGroup(BIOSIG).Touchlessfingerprintsamplequality:Prerequisitesfortheapplicabilityofnfiq2.0(IEEENewYork,2020),pp.1–5. GoogleScholar  B.Yang,G.Li,C.Busch,in20thInternationalConferenceonImageProcessing(ICIP).Qualifyingfingerprintsamplescapturedbysmartphonecameras(IEEENewYork,2013),pp.4161–4165. GoogleScholar  G.Li,B.Yang,C.Busch,QualifyingFingerprintSamplesCapturedbySmartphoneCamerasinReal-LifeScenarios(2016).https://ntnuopen.ntnu.no/ntnu-xmlui/handle/11250/2388306.Accessed15Feb2021.K.Ko,User’sguidetoNISTbiometricimagesoftware(NBIS).Techrep,NISTInteragency/InternalRep(NISTIR)-7392(2007).R.D.Labati,A.Genovese,V.Piuri,F.Scotti,etal,TowardUnconstrainedFingerprintRecognition:AFullyTouchless3-DSystemBasedonTwoViewsontheMove,.Trans.Syst.ManCybern.Syst.46(2),202–219(2016).Article  GoogleScholar  Y.Tang,F.Gao,J.Feng,Y.Liu,inInternationalJointConferenceonBiometrics(IJCB).FingerNet:Anunifieddeepnetworkforfingerprintminutiaeextraction(IEEENewYork,2017),pp.108–116. GoogleScholar  F.Han,J.Hu,M.Alkhathami,K.Xi,in6thConferenceonIndustrialElectronicsandApplications(ICIEA).Compatibilityofphotographedimageswithtouch-basedfingerprintverificationsoftware(IEEENewYork,2011),pp.1034–1039. GoogleScholar  M.D.Garris,C.I.Watson,R.McCabe,C.L.Wilson,User’sguidetonistfingerprintimagesoftware(nfis).Tech.rep.,NISTInteragency/InternalRep.(NISTIR)-6813(2001).K.Wang,J.Jiang,Y.Cao,X.Xing,R.Zhang,inPatternRecognition.Preprocessingalgorithmresearchoftouchlessfingerprintfeatureextractionandmatching(SpringerSingapore,2016),pp.436–450.Chapter  GoogleScholar  S.S.Agaian,M.M.A.Mulawka,R.Rajendran,S.P.Rao,S.K.KM,S.Rajeev,Acomparativestudyofimagefeaturedetectionandmatchingalgorithmsfortouchlessfingerprintsystems.Electron.Imaging.2016(15),1–9(2016). GoogleScholar  B.Y.Hiew,A.B.J.Teoh,O.S.Yin,Asecuredigitalcamerabasedfingerprintverificationsystem.J.Vis.Commun.ImageRepresent.21(3),219–231(2010).Article  GoogleScholar  S.Mallat,Groupinvariantscattering.Commun.Pur.Appl.Math.65(10),1331–1398(2012).MathSciNet  MATH  Article  GoogleScholar  NeuroTechnology,VerifingerSDK(2020).https://doi.org/www.neurotechnology.com/verifinger.html.R.Cappelli,M.Ferrara,D.Maltoni,Minutiacylinder-code:Anewrepresentationandmatchingtechniqueforfingerprintrecognition.Trans.PatternAnal.Mach.Intell.32(12),2128–2141(2010).Article  GoogleScholar  X.Yin,Y.Zhu,J.Hu,ContactlessFingerprintRecognitionBasedonGlobalMinutiaTopologyandLooseGeneticAlgorithm.Trans.Inf.ForensicsSecur.15:,28–41(2020).Article  GoogleScholar  R.Vyas,A.Kumar,ACollaborativeApproachusingRidge-ValleyMinutiaeforMoreAccurateContactlessFingerprintIdentification.arXiv:1909.06045[cs,eess](2019).R.D.Labati,A.Genovese,E.Muñoz,V.Piuri,F.Scotti,Anovelporeextractionmethodforheterogeneousfingerprintimagesusingconvolutionalneuralnetworks.PatternRecogn.Lett.113:,58–66(2017).Article  GoogleScholar  A.Kumar,C.Kwong,inConferenceonComputerVisionandPatternRecognition(CVPR).TowardsContactless,Low-CostandAccurate3dFingerprintIdentification(IEEENewYork,2013),pp.3438–3443. GoogleScholar  Q.Zheng,A.Kumar,G.Pan,inInternationalWorkshoponBiometricsandForensics(IWBF).Contactless3dfingerprintidentificationwithout3dreconstruction(IEEENewYork,2018),pp.1–6. GoogleScholar  S.Orandi,J.Libert,B.Bandini,K.Ko,J.Grantham,C.Watson,Evaluatingtheoperationalimpactofcontactlessfingerprintimageryonmatcherperformance.Tech.Rep.NISTIR8315,NatlInst.Stand.Technol.(2020).A.Lindoso,L.Entrena,J.Liu-Jimenez,E.SanMillan,inAdvancesinBiometrics.Correlation-basedfingerprintmatchingwithorientationfieldalignment(Springer-VerlagBerlin,2007),pp.713–721.Chapter  GoogleScholar  R.D.Labati,V.Piuri,F.Scotti,inWorkshoponComputationalIntelligenceinBiometricsandIdentityManagement(CIBIM).Aneural-basedminutiaepairidentificationmethodfortouch-lessfingerprintimages(IEEENewYork,2011),pp.96–102. GoogleScholar  A.Malhotra,A.Sankaran,M.Vatsa,R.Singh,OnMatchingFinger-SelfiesUsingDeepScatteringNetworks.Trans.Biom.Behav.IdentitySci.2(4),350–362(2020).Article  GoogleScholar  C.Lin,A.Kumar,ACNN-BasedFrameworkforComparisonofContactlesstoContact-BasedFingerprints.Trans.Inf.ForensicsSecur.14(3),662–676(2019).Article  GoogleScholar  H.Tan,A.Kumar,TowardsMoreAccurateContactlessFingerprintMinutiaeExtractionandPose-InvariantMatching.Trans.Inf.ForensicsSecur.15:,3924–3937(2020). GoogleScholar  J.Galbally,G.Bostrom,L.Beslay,inInternationalJointConferenceonBiometrics(IJCB).Full3dtouchlessfingerprintrecognition:Sensor,databaseandbaselineperformance(IEEENewYork,2017),pp.225–233. GoogleScholar  S.M.Furman,B.C.Stanton,M.F.Theofanos,J.M.Libert,J.D.Grantham,Contactlessfingerprintdevicesusabilitytest.Tech.Rep.NISTIR8171,NatlInst.Stand.Technol.(2017).A.Pillai,S.Mil’shtein,inConferenceonTechnologiesforHomelandSecurity(HST).Cancontactlessfingerprintsbecomparedtoexistingdatabase?(IEEENewYork,2012),pp.390–394. GoogleScholar  ISO,ISO/IEC19794-4:2011:Informationtechnology–Biometricdatainterchangeformats–Part4:Fingerimagedata,vol.2011(InternationalOrganizationforStandardization,Geneva,2011). GoogleScholar  P.Wild,F.Daubner,H.Penz,G.F.Domínguez,in7thInternationalWorkshoponBiometricsandForensics(IWBF).ComparativeTestofSmartphoneFingerPhotovs.Touch-basedCross-sensorFingerprintRecognition(IEEENewYork,2019),pp.1–6. GoogleScholar  A.Dabouei,S.Soleymani,J.Dawson,N.M.Nasrabadi,DeepContactlessFingerprintUnwarping,1–8(2020).J.Libert,J.Grantham,B.Bandini,K.Ko,S.Orandi,C.Watson,InteroperabilityAssessment2019:Contactless-to-ContactFingerprintCapture.Tech.rep.NatlInst.Stand.Technol.(2020).C.Sousedik,C.Busch,Presentationattackdetectionmethodsforfingerprintrecognitionsystems:asurvey.IETBiom.3(4),219–233(2014).Article  GoogleScholar  Y.S.Moon,J.S.Chen,K.C.Chan,K.So,K.C.Woo,Waveletbasedfingerprintlivenessdetection.Electron.Lett.41(20),1112–1113(2005).Article  GoogleScholar  N.Overgaard,C.Sousedik,C.Busch,EulerianVideoMagnificationforFingerprintLivenessDetection.NISKJ.(2014).A.Taneja,A.Tayal,A.Malhorta,A.Sankaran,M.Vatsa,R.Singh,in8thInternationalConferenceonBiometricsTheory,ApplicationsandSystems(BTAS).Fingerphotospoofinginmobiledevices:Apreliminarystudy(IEEENewYork,2016),pp.1–7. GoogleScholar  C.Rathgeb,A.Uhl,Asurveyonbiometriccryptosystemsandcancelablebiometrics.EURASIPJ.Inf.Secur.2011(3)(2011).K.Nandakumar,A.K.Jain,Biometrictemplateprotection:Bridgingtheperformancegapbetweentheoryandpractice.SignalProc.Mag.Spec.IssueBiom.Secur.Priv.32(5),88–100(2015).Article  GoogleScholar  ISO/IECJTC1SC27SecurityTechniques,ISO/IEC24745:2011.InformationTechnology-SecurityTechniques-BiometricInformationProtection(2011).A.Juels,M.Sudan,Afuzzyvaultscheme.Des.CodesCrypt.38(2),237–257(2006).MathSciNet  MATH  Article  GoogleScholar  K.Nandakumar,A.K.Jain,S.Pankanti,Fingerprint-basedfuzzyvault:Implementationandperformance.IEEETrans.Inf.ForensicsSecur.2(4),744–757(2007).Article  GoogleScholar  A.Juels,M.Wattenberg,in6thACMConferenceonComputerandCommunicationsSecurity.Afuzzycommitmentscheme(AssociationforComputingMachineryNewYork,1999),pp.28–36. GoogleScholar  A.B.J.Teoh,J.Kim,Securebiometrictemplateprotectioninfuzzycommitmentscheme.IEICEElectron.Express.4(23),724–730(2007).Article  GoogleScholar  N.K.Ratha,J.H.Connell,R.M.Bolle,Enhancingsecurityandprivacyinbiometrics-basedauthenticationsystems.IBMSyst.J.40(3),614–634(2001).Article  GoogleScholar  N.Ratha,J.Connell,R.M.Bolle,S.Chikkerur,in18thInternationalConferenceonPatternRecognition(ICPR),4.Cancelablebiometrics:Acasestudyinfingerprints(IEEENewYork,2006),pp.370–373. GoogleScholar  M.-H.Lim,A.B.J.Teoh,J.Kim,Biometricfeature-typetransformation:Makingtemplatescompatiblefortemplateprotection.SignalProc.Mag.32(5)(2015).S.B.Zannou,T.Djara,A.Vianou,in3rdInternationalConferenceonBio-engineeringforSmartTechnologies(BioSMART).Securedrevocablecontactlessfingerprinttemplatebasedoncenterofmass(IEEENewYork,2019),pp.1–4. GoogleScholar  Q.Lai,Z.Wan,A.Akgul,O.F.Boyraz,M.Z.Yildiz,Designandimplementationofanewmemristivechaoticsystemwithapplicationintouchlessfingerprintencryption.ChineseJ.Phys.67:,615–630(2020).MathSciNet  Article  GoogleScholar  A.Malhotra,S.Chhabra,M.Vatsa,R.Singh,inConferenceonComputerVisionandPatternRecognitionWorkshops(CVPRW).Onprivacypreservinganonymizationoffinger-selfies(IEEENewYork,2020),pp.26–27. GoogleScholar  A.Ross,K.Nandakumar,A.K.Jain,HandbookofMultibiometrics,vol.6(SpringerScience&BusinessMedia,Boston,2006). GoogleScholar  D.Noh,W.Lee,B.Son,J.Kim,Empiricalstudyontouchlessfingerprintrecognitionusingaphonecamera.J.Electron.Imaging.27(3),033038(2018). GoogleScholar  A.K.Jain,B.Klare,A.A.Ross,in8thInternationalConfonferenceonBiometrics(ICB).Guidelinesforbestpracticesinbiometricsresearch(IEEENewYork,2015),pp.1–5. GoogleScholar  W.Zhou,J.Hu,I.Petersen,S.Wang,M.Bennamoun,in11thInternationalConferenceonFuzzySystemsandKnowledgeDiscovery(FSKD).Abenchmark3dfingerprintdatabase(IEEENewYork,2014),pp.935–940. GoogleScholar  A.Kumar,TheHongKongPolytechnicUniversityContactless2DtoContact-based2DFingerprintImagesDatabaseVersion1.0(2017).http://www4.comp.polyu.edu.hk/~csajaykr/fingerprint.htm.Accessed15Feb2021.AjayKumar,TheHongKongPolytechnicUniversityFingerImageDatabaseVersion1.0(2010).http://www4.comp.polyu.edu.hk/~csajaykr/fvdatabase.htm.Accessed15Feb2021.A.Kumar,TheHongKongPolytechnicUniversityLowResolutionFingerprintDatabase,Version1.0(2011).http://www4.comp.polyu.edu.hk/~csajaykr/fplr.htm.Accessed15Feb2021.S.Chopra,A.Malhotra,M.Vatsa,R.Singh,inProceedingsoftheIEEEConferenceonComputerVisionandPatternRecognitionWorkshops(CVPRW).UnconstrainedFingerphotoDatabase(IEEENewYork,2018),pp.517–525. GoogleScholar  D.Gafurov,P.Bours,B.Yang,C.Busch,inInternationalConferenceonComputationalScienceandItsApplications(ICCSA).Guc100multi-scannerfingerprintdatabaseforin-house(semi-public)performanceandinteroperabilityevaluation(IEEENewYork,2010),pp.303–306. GoogleScholar  DownloadreferencesAcknowledgementsNotapplicable.FundingTheauthorsacknowledgethefinancialsupportbytheFederalMinistryofEducationandResearchofGermanyintheframeworkofMEDIAN(FKZ13N14798).ThisresearchworkhasbeenpartiallyfundedbytheGermanFederalMinistryofEducationandResearchandtheHessianMinistryofHigherEducation,Research,ScienceandtheArtswithintheirjointsupportoftheNationalResearchCenterforAppliedCybersecurityATHENE.AuthorinformationAuthorsandAffiliationsda/sec–BiometricsandInternetSecurityResearchGroup,HochschuleDarmstadt,Schöfferstraße8b,Darmstadt,64295,GermanyJannisPriesnitz, ChristianRathgeb & ChristophBuschFreieUniversitätBerlin,Germany,Takustraße9,Berlin,14195,GermanyNicolasBuchmann & MarianMargrafAuthorsJannisPriesnitzViewauthorpublicationsYoucanalsosearchforthisauthorin PubMed GoogleScholarChristianRathgebViewauthorpublicationsYoucanalsosearchforthisauthorin PubMed GoogleScholarNicolasBuchmannViewauthorpublicationsYoucanalsosearchforthisauthorin PubMed GoogleScholarChristophBuschViewauthorpublicationsYoucanalsosearchforthisauthorin PubMed GoogleScholarMarianMargrafViewauthorpublicationsYoucanalsosearchforthisauthorin PubMed GoogleScholarContributionsThemaincontributionwasdonebyJannisPriesnitzandChristianRathgeb.Allotherauthorssupportedthetechnicalandscientificwork.CorrespondingauthorCorrespondenceto JannisPriesnitz.Ethicsdeclarations Competinginterests Theauthorsdeclarethattheyhavenocompetinginterests. AdditionalinformationPublisher’sNoteSpringerNatureremainsneutralwithregardtojurisdictionalclaimsinpublishedmapsandinstitutionalaffiliations.Rightsandpermissions OpenAccessThisarticleislicensedunderaCreativeCommonsAttribution4.0InternationalLicense,whichpermitsuse,sharing,adaptation,distributionandreproductioninanymediumorformat,aslongasyougiveappropriatecredittotheoriginalauthor(s)andthesource,providealinktotheCreativeCommonslicence,andindicateifchangesweremade.Theimagesorotherthirdpartymaterialinthisarticleareincludedinthearticle’sCreativeCommonslicence,unlessindicatedotherwiseinacreditlinetothematerial.Ifmaterialisnotincludedinthearticle’sCreativeCommonslicenceandyourintendeduseisnotpermittedbystatutoryregulationorexceedsthepermitteduse,youwillneedtoobtainpermissiondirectlyfromthecopyrightholder.Toviewacopyofthislicence,visithttp://creativecommons.org/licenses/by/4.0/. ReprintsandPermissionsAboutthisarticleCitethisarticlePriesnitz,J.,Rathgeb,C.,Buchmann,N.etal.Anoverviewoftouchless2Dfingerprintrecognition. JImageVideoProc.2021,8(2021).https://doi.org/10.1186/s13640-021-00548-4DownloadcitationReceived:26March2020Accepted:02February2021Published:24February2021DOI:https://doi.org/10.1186/s13640-021-00548-4SharethisarticleAnyoneyousharethefollowinglinkwithwillbeabletoreadthiscontent:GetshareablelinkSorry,ashareablelinkisnotcurrentlyavailableforthisarticle.Copytoclipboard ProvidedbytheSpringerNatureSharedItcontent-sharinginitiative KeywordsBiometricsFingerprintrecognitionTouchlessContactlessFingerimageFingerphoto DownloadPDF Advertisement