Extending the imaging volume for biometriciris recognitionRamkumar Narayanswamy,Gregory E.Johnson,Paulo E.X.Silveira,and Hans B.WachThe use of the human iris as a biometric has recently attracted significant interest in the area of securityapplications.The need to capture an iris without active user cooperation places demands on the opticalsystem.Unlike a traditional optical design,in which a large imaging volume is traded off for diminishedimaging resolution and capacity for collecting light,Wavefront Coded imagingis a computational imagingtechnology capable of expanding the imaging volume while maintaining an accurate and robust irisidentification capability.We apply Wavefront Coded imaging to extend the imaging volume of the irisrecognition application.2005 Optical Society of America0 CIS codes:.100.5010,100.2980,170.0110.1.Introductionof a commercial implementation).In this applicationRecent interest in the use of biometric identificationan iris recognition camera is typically mounted ap-for security applications has led to the search forproximately 22 in.(1 in.2.54 cm)from the user.alternatives that are less error prone and less intru-The iris system validates the user at the start of eachsive than traditional fingerprint identification.session and may continue to validate at regular in-Among the alternatives,iris recognition is particu-tervals.Some current commercial iris imaging sys-larly attractive owing to the high degree of entropytems,which operate at F/8,have an imaging volumeper unit area encountered in human irises,1 com-of 3 in3.(1 in.X 1.5 in.X 2 in.).This imaging vol-bined with the extremely small rate of change inume is restrictive and not conducive for easy use.Thehuman iris patterns with age and health conditions.user must be trained to use these types of systemsFor example,identical twins have completely distinctand must actively cooperate every time his or her irisirises.The left-eye iris of a person bears no correla-is to be validated.It is expected that increasing thetion with the right-eye iris.However,dynamic irisimaging volume to6in.×6in.×10in,a factor-.recognition presents some formidable challenges toof-100 increase over the traditional imaging system,the optical designer.To capture the iris informationwill make iris biometric recognition easier to use andat the fidelity necessary for reliable identification,themore commercially acceptable.However,this in-optical system must maintain a high resolution overcrease in imaging volume must be achieved whilethe entire field of view and depth of focus.Moreover,maintaining recognition and rejection accuracy,for robustness and ubiquity,the system should reli-which translates to the optical requirements of highably identify subjects over an extended field of viewimaging resolution and low F-numbers.The low-Fand depth of focus.number requirement is also important for keepingIris images as biometrics are currently being usedthe illumination intensity low enough to ensure eyefor computer security (see Fig.1 for an schematicsafety under all conditions.These imaging-systemexample,or Panasonic's Authenticam for an examplecharacteristics are summarized in Table 1.The field of view and the depth of field trade-offsrequired in traditional imaging systems can lead toThe authors are with CDM Optics,Ine.,4001 Discovery Drive,impractical system designs.In traditional optical sys-Suite 130,Boulder,Colorado 80303-7816.R.Narayanswamy'stems,increasing the field of view dictates that thee-mail address is ramkumarn@cdm-optics.com.imaging optics become significantly more complex,Received 9 June 2004;revised manuscript received 13 Octobermaking it increasingly harder to control the aberra-2004:accepted 14 October 2004.tions that arise from imaging over large angular re-0003-6935/05/050701-12$15.00/0gions.The depth of field is typically increased by2005 Optical Society of Americareducing the system aperture.However,smaller ap-10 February 2005/Vol.44,No.5/APPLIED OPTICS701Wavefront Coded ImagingOptical SystemProcessingApplicationWavefront CodedProcessingAspheric OpticalDigitalElementFig.2.Architecture of a computational imaging system consist-ing of application-specific optics and complementary signal pro-cessing.The optics and signal processing are jointly optimized fora particular imaging application.Fig.1.Iris recognition system for computer security.An ideal irisoptical systems.Section 3 describes an iris recogni-recognition system captures and recognizes the iris accurately overtion system and the system-design challenges thata large imaging volume and requires no active cooperation frompertain to iris identification.Section 4 uses an off-the user.the-shelf Wavefront Coded iris recognition system toexperimentally show the benefits it offers over a tra-ertures reduce the overall system resolution,whichditional system for iris recognition.Section 5 explainsleads to loss of potentially valuable information in thein detail the design of an application-specific Wave-higher spatial frequencies.Furthermore,a smallerfront Coded optical system that is optimized for anaperture reduces the light-gathering capacity of theiris recognition algorithm and presents the resultssystem.Reduction in light-gathering capacity is nor-expected when an optimized design is used.Section 6mally addressed by the use of longer integration pe-offers concluding remarks and points to areas of fu-ture research.riods or higher illumination levels.The former canlead to motion blur and the latter to eye-safety issueswith active illumination.Thus the traditional ap-Wavefront Coded Imagingproach leaves the optical system designer no otherWavefront Coded imaging is a novel imaging para-choice but to trade off resolution and light-gatheringdigm in which the optics and their complementarycapacity for an increased depth of field.signal processing comprise an inherent aspect of im-Wavefront Coded imaging,a computational imag-aging.23 General Wavefront Coded imaging systemsing method,can deliver a large field of view and depthtypically consist of specialized Wavefront Coded as-of field without increasing the number of optical ele-pherical optics,a digital detector and a decoding orments in the design or sacrificing the light-capturingprocessing step,as shown in Fig.2.capacity.Wavefront Coded systemsconsist ofIn iris imaging the optics are focused to the nomi-application-specific aspheric optical surfaces,fol-nal object distance,and images of the scene are re-lowed by signal processing of the captured imagescorded exactly as in a traditional system.TheThe specialized optics act as an encoder of the imagedetected image is processed with a decoding filter,information,and the signal processing acts as a de-which produces the decoded image.The decoding fil-coder.The disadvantages of this type of imaging areter can be implemented through convolution,oftena reduction in the signal-to-noise ratio(SNR)at bestderived from the point-spread function(PSF)of thefocus and the need to digitally process the acquiredWavefront Coded imaging system.If the image re-images.The former,as will be shown in this paper,iscorded by the detector is examined before processing,quite acceptable for biometric iris recognition,it will appear as a blurred version of the eye.But thewhereas the latter imposes a very small processingblur will be uniform across the image and will notoverhead in applications that require digital process-vary as a function of field or object distances.Thising of the images,as is the case in iris recognition.blurred image is an image optimized for informationThis paper is organized as follows.Section 2 pro-capture,as opposed to human visualization.vides a short introduction to the use of WavefrontFigure 3 shows a comparison between the modula-Coded optics for extending the imaging volume oftion transfer function(MTF)as a function of normal-ized spatial frequency (where unity represents thecutoff spatial frequency of the optical system)of aTable 1.Characteristics of an Ideal Imaging System fortraditional system and a Wavefront Coded opticalIris Recognitionsystem at best focus(solid curve)and at an amount ofCharacteristicsdefocus(dotted curve)arbitrarily selected but equallylarge in both plots.Note that,although the tradi-High iris recognition and rejection accuracy.tional (diffraction limited)optical system presentsEase of use-minimal user cooperation requiredthe largest MTF at all spatial frequencies [Fig.3(a)],Large image-capture volume.the MTF quickly degrades as a function of defocusHigh light-capture capacity (low F/#,short exposure period).and presents nulls at multiple frequencies,repre-702APPLIED OPTICS Vol.44,No.5 /10 February 2005Best Focus+DefocusedSclera2400.10.20.30.40.50.60.7Normalized Spatial FrequencyBest FocusDefocused10-15Fig.4.Iris texture as a biometric.The rich texture of the irisdiffers significantly even between the left and the right eyes of the0.1020.30.40.50.60.70.80.9same person.This texture is encoded as a 2048-bit vector thatNormalized Spatial Frequencyforms the basis of the biometric.Fig.3.Comparison between the modulation transfer function of(a)a traditional optical system and (b)a Wavefront Coded opticalsystem.Notice thatthe traditional system has multiple nulls in theorder of 1 in 1010.This false acceptance rate,which isMTF,whereas the Wavefront Coded system maintains the systema function of the decision threshold,can be changed toalmost invariant with defocus.The same amount of defocus issuit the specific application need,as described byapplied in both cases.Daugman.1 Under realistic conditions,motion blurand noise may deteriorate the quality of the imagessenting the irremediable loss of information.In con-available for comparison.Under these conditions,thethreshold level may be reduced to reduce the falsetrast,although the Wavefront Coded system [Fig.3(b)]presents a lower MTF at best focus,it is capablerejection rate at the expense of an increase in theof maintaining a nearly invariable MTF over a widefalse acceptance rate.range of defocus.Moreover,no nulls are present inThe use of the human iris as a means of identifi-cation has many advantages over other biometrics.the intermediate spatial frequencies,allowing for fullrecovery of the image information after processing.Unlike fingerprints and the human face and voice,Wavefront Coding has been shown to effectivelythe iris remains unchanged over the majority of aincrease the depth offocus of numerous digital opticalperson's life.From an operational perspective,theiris texture is attractive as a biometric owing to thesystems4-10 and has been used for reducing the costsand increasing the robustness of optical systems bypossibility of imaging it from a distance:Cooperativeincreasing their aberration tolerance.11-14 Often,theusers can be conveniently identified without muchprimary consideration of these applications has beeneffort,and uncooperative users may be identifiedthe visual quality of the images acquired,which is awithout their knowledge.Furthermore,the pupil hasa physiological response to light,and the pupil's rapidsubjective parameter.In this paper we demonstratethe benefits of using Wavefront Coded imaging in irisdiameter flutter acts as a natural test against artifice.recognition,which represents a class of applicationsAFunctional Blocks of the Iris Recognition Systemin which the performance metrics are quantitative.allowing for a nonsubjective analysis of the trade-offsThe iris recognition system can be functionally sep-of and advantages provided by Wavefront Coding.arated into the following blocks:i.An infrared illumination system,which enables3.Iris Recognition Systemthe imaging of the iris texture with good contrast.The visible part of the human eye consists of theThe illumination power level should be safe to thepupil,the iris,and the sclera,as shown in Fig.4.Theeye,but high enough to support short exposure peri-color of the iris is of little use in recognition,but theods.texture of the iris is quite complex and unique.Evenii.An imaging system,which acquires the imagedeeply pigmented irises,which appear black at visi-electronically for biometric processing.Ideally,theble wavelengths,show a rich texture with near-image should be free of defocus and aberration errorsinfrared (NIR)illumination.This rich iris texture hasover a wide range of object distances and field posi-a significantly unique signature such that the oper-tions and should facilitate reliable iris-score genera-ating probability of false acceptance can be of thetion.10 February 2005 Vol.44,No.5 APPLIED OPTICS703