WO2009029638A1 - Iris recognition - Google Patents

Abstract

Improved system and methods are provided for iris recognition.

Description

IRIS RECOGNITION

CROSS-REFERENCE TO RELATED APPLICATION

[0001] This application claims the priority benefit of U.S. Provisional Application Serial No. 60/968,271 , filed on August 27, 2007, pending, which is hereby incorporated herein by reference in their entirety.

BACKGROUND OF THE INVENTION

[0002] This invention relates to iris recognition_* specifically for human identification and identity verification.

Iris images can be hard segment and process for identification. Image analysis algorithms preprocess, locate, and extract the iris structure from a digital image of the human eye.

[0003] Iris recognition is a method of biometric authentication that uses pattern recognition techniques based on images of the irides of an individual's eyes. Iris recognition uses camera technology, and subtle illumination to reduce specular reflection from the convex cornea to create images of the detail-rich, intricate structures of the iris. These unique structures converted into digital templates, provide mathematical representations of the iris that yield unambiguous positive identification of an individual.

[0004] The iris of the eye has been described as the ideal part of the human body for biometric identification for several reasons. It is an internal organ that is well protected against damage and wear by a highly transparent and sensitive membrane (the cornea). This distinguishes it from fingerprints, which can be difficult to recognize after years of certain types of manual labor. The iris is mostly flat and its geometric configuration is only controlled by two complementary muscles (the sphincter pupillae and dilator pupillae), which control the diameter of the pupil. This makes the iris shape far more predictable than, for instance, that of the face. The iris has a fine texture that - like fingerprints - is determined randomly during embryonic gestation. Even genetically identical individuals have completely independent iris textures, whereas DNA (genetic "fingerprinting") is not unique for the about 1.5% of the human population who have a genetically identical monozygotic twin. While there are some medical and surgical procedures that can affect the colour and overall shape of the iris, the fine texture remains remarkably stable over many decades. Some iris identification have succeeded over a period of about 30 years. [0005] An iris scan is similar to taking a photograph and can be performed from about 10 cm to a few meters away. There is no need for the person to be identified to touch any equipment that has recently been touched by a stranger, thereby eliminating an objection that has been raised in some cultures against finger-print scanners, where a finger has to touch a surface, or retinal scanning, where the eye can be brought very close to a lens (like looking into a microscope lens). Moreover, it may be that a focused digital photograph with an Ms diameter of about 200 pixels, for example, contains much more long-term stable information than a fingerprint. [0006] However, there remains a need to develop systems and methods to obtain positive iris recognition more rapidly and with more robust accuracy.

SUMMARY OF THE INVENTION

[0007] In general, the methods and system of the invention are directed to identification of a subject using iris recognition. In one aspect of the invention, an iris recognition method for uniquely identifying a particular human being by biometric analysis of the iris of the eye, comprising preprocessing an acquired image of an eye of said human to be identified; obtaining pupil segmentation information; obtaining iris segmentation information; normalizing said pupil segmentation and iris segmentation information to obtain a pattern recognition code; and comparing said pattern recognition code with a reference iris code to determine to uniquely identify said particular human being if said pattern recognition and reference iris code match. Any of the steps herein can be performed using computer program product that comprises a computer executable logic recorded on a computer readable medium.

[0008] In various embodiments, preprocessing comprises at least removing reflection corresponding to an illumination source from said acquired image; computing a global binary threshold corresponding to said acquired image in order to convert an intensity image of the eye into a binary coded image. In one embodiment, computer executable logic is provided that conducts functions such as obtaining an image, preprocessing an image, removal of reflections, performing computations, comparing a test image with a reference image, or other functions described herein.

[0009] In further embodiments, iris recognition methods and systems of the invention further comprise determining a morphological opening of said threshold to obtain an opened image that isolates the reflection from the noise in the image. [0010] In some embodiments, methods and systems of the invention further comprise identifying one or more holes in said opened image and/or selecting said one or more holes and filling said holes with an approximate value from surrounding pixels. In one embodiment, computer executable logic is provided that conducts morphological opening of an iris image (to produce an 'opened image'), and isolates reflections present in the image from noise background signals present in the image. Furthermore, such computer executable logic conducts identification of such one or more holes in the opened image.

[0011] Pupil segmentation can comprise segmenting said pupil from said iris at the pupillary boundary. In a further embodiment, the iris image is converted to or provided in grayscale. In one embodiment, a computer executable logic conducts conversion of the iris image (e.g., digital color photograph) to a grayscale image. [0012] In one embodiment, methods or systems for iris recognition further comprise computing a minimum intensity of said grayscale iris image and converting said image to binary code based on said minimum intensity. In one embodiment, such computations are conducted by a computable executable logic of the invention. [0013] In one embodiment, methods or systems for iris recognition further comprises performing a pupil detection error check. In one example, the pupil error check comprises computing Euclidean Distance Transform (EDT) from said grayscale image to the nearest nonzero pixel and converting said iris image to binary code using said EDT.

[0014] In an additional embodiment, a step of eyelash detection is performed. Eyelash detection can comprise detecting edges of said grayscale iris image. In addition, eyelash detection can comprise filtering and dilating said edges to detect said eyelash. Furthermore, methods or systems for iris recognition may comprise removing the eyelash structure from the image. [0015] In another embodiment, methods or systems for iris recognition further comprise removing spurious objects and filling said holes. For example, in one embodiment, filling comprises binary hole filling. [0016] In a further embodiment, removing spurious objects comprises binary filtering and/or pupil selection.

[0017] In one embodiment, pupil selection comprises selecting the largest circular binary object in said image. [0018] If a pupil is not detected, a minimum intensity is scaled up until a binary object is detected corresponding to a certain pupil area. [0019] In one embodiment, methods or systems for iris recognition comprises morphological closing and hole filling to complete a pupil structure and detecting the largest circular object (e.g., during pupil segmentation). In additional embodiments, measurements are obtained for the center coordinates and horizontal radius corresponding to said pupil. [0020] In one embodiment, iris segmentation comprises segmenting said iris from the sclera at the limbic boundary of said iris image and/or selecting iris pixel intensity. Furthermore, selecting iris pixel intensity can comprise defining regions of interest outside of east and west boundaries of said pupil segmentation. In additional embodiments, the region of interest is converted into a mean intensity signal and measuring the horizontal sampled derivative of said regions of interest. [0021] In an additional embodiment, methods and systems of the invention further comprise identifying a maximum peak, defining a localized region of interest based on an approximate limbic boundary and computing a maximum, mean and minimum intensities within said localized region of interest.

[0022} Such maximum and minimum intensities can be used to obtain a binary of said grayscale iris image.

[00231 Another embodiment of the invention, methods and systems for iris recognition comprise a step for performing a circular segmentation. Such circular segmentation comprises measuring a circular boundary around said iris based on a limbic boundary approximation and removing one or more features outside said boundary. [0024] In one embodiment, methods and systems of the invention comprise iris segmentation comprising removing reaming spurious features and holes in said iris image. Such iris segmentation can comprise detecting a largest circular object boundary. Furthermore, such methods can comprise measuring center coordinates and horizontal radius corresponding to said iris.

[0025} In various embodiments, methods and systems of the invention comprise a normalization comprising conversion of said iris segmentation to a rectangular orientation. In further embodiments, a radius for said iris is measured and/or of claim a rectangular orientation is configured based on measurements of center and horizontal coordinates for the iris and the pupil. [0026] In one embodiment, based on center coordinates of said iris, center coordinates of said pupil and radius for said iris a conversion to a rectangular orientation is made.

[0027) Additionally, a step of performing image enhancement can be included, which can comprise filtering an unwrapped image with an averaging filter, subtracting a filtered image from an original unwrapped image to enhance structural content and minimum and maximum normalization of image intensities. Performing image enhancement can improve the accuracy and speed for iris recognition.

[0028] In a further embodiment, a template mask can be generated. For example, such generation is obtained by blocking features not related to said iris. In an additional embodiment, wavelet filtering of said iris image can be performed to optimize image enhancement. [0029] In another embodiment, structural features in an image are converted into a binary template. Furthermore, hamming distance(s) can be measured using such a binary template.

[0030] In one embodiment, a method for iris recognition is provided, comprising the steps of acquiring an image of at least one eye of a user, and of processing said image to remove reflection, conversion of said image to grayscale to measure minimum and maximum intensities, determining the outer boundary of the iris from said minimum and maximum intensities, filling in holes with intensities from surrounding pixel intensities; and determining if the processed image matches to a reference image. [0031] In another embodiment, a method for iris recognition is provided comprising: providing an image of an eye; selecting a pupil in the image; segmenting the pupil; selecting an iris in the image; segmenting the iris; wherein said segmenting detects and removes reflections in said image to enhance said segmentation; and determining if said image matches to a reference image. [0032] In various methods of the invention for iris recognition a quality metric is obtained. For example, a quality metric can be based on image quality parameters of pupil segmentation, occlusion of the iris, size of the pupil dilation, pupil constriction, number of pixels inside the iris and clarity of iris pixels. For example, the quality metric is a score (0-100) calculated based on the following features of the algorithm and image: blurriness/noisiness of the image, pupil segmentation circularity, iris segmentation circularity, iris pixel resolution, and occlusion estimates.

[0033] In various other embodiments of the invention, segmentation comprises converting said image into a binary coded image. For example, converting is before or after a step comprising determining a pupil or limbic boundary.

[0034] In one aspect of the invention, an iris recognition system is provided for biometric identification implemented through a process of dynamic thresholding, binary conversions, and morphological operations. For example an image of a subject's eye is provided and processed to confirm identity of the subject. In various embodiments, light reflections that are present in the image may be isolated and removed in order to facilitate recognition processing. In some embodiments, the pupil of the eye image may be detected and isolated from the image. Furthermore, a circular approximation may be fit to the eye using an estimated radius of the isolated pupil. An approximate iris boundary radius may be detected by deriving an intensity signal of the eye image pixels on the east and west side of the pupil. The iris of the eye image may be isolated by defining regions of interest containing iris pixels and dynamically binary thresholding the eye image. Once the iris is separated from portions of the eye image, the iris may be normalized and its structural composition may be enhanced and/or masked. A quality factor may be computed from factors of the original eye image, iris segmentation, enhanced image, or masked image in order to evaluated image matchability. The iris may be encoded and analyzed.

INCORPORATION BY REFERENCE

[0035] All publications and patent applications mentioned in this specification are herein incorporated by reference to the same extent as if each individual publication or patent application was specifically and individually indicated to be incorporated by reference.

BRIEF DESCRIPTION OF THE DRAWINGS

[0036] The novel features of the invention are set forth with particularity in the appended claims. A better understanding of the features and advantages of the present invention will be obtained by reference to the following detailed description that sets forth illustrative embodiments, in which the principles of the invention are utilized, and the accompanying drawings of which: [0037] Figure 1 illustrates iris structure enhancement and encoding binary encoded image of the enhanced iris structure.

[0038] Figure 2 illustrates a graph for Hamming distance distribution of genuine and imposter scores of approximately 1 million iris comparisons. [0039] Figure 3 illustrates iris boundary detection and segmentation: Left image: actual iris segmentation boundary, Right image: best approximate circular fit iris segmentation boundary. [0040] Figure 4 illustrates pupil boundary detection and segmentation: Left image: actual pupil segmentation boundary, Right image: best approximate circular fit pupil segmentation boundary.

[0041] Figure 5 provides a quality score v. EER graph: An example of the equal error rate (the point where the genuine and imposter distributions overlap and have the same value on a Receiver Operating Characteristic Curve) versus the derived quality metric of approximately 1 million iris comparisons.

[0042] Figure 6 illustrates a quality score v. accept rate: Genuine accept rate versus the derived quality metric of approximately 1 million iris comparisons.

[0043] Figure 7 provides graph characterizing Equal Error Rate (EER).

[0044] Figure 8 illustrates reflection removal: Left image: original iris image, Right image: Removal of large light source reflections from any portion of the image.

[0045] Figure 9 illustrates circular boundary projected around the grayscale iris image iris segmentation.

[0046] Figure 10 illustrates edge detection of all edges in a grayscale iris image.

[0047] Figure 11 illustrates binary image of detected eyelash portions of a grayscale iris image after edge detection filtering and dilatation. [0048] Figure 12 illustrates an iris binary coded image, binary conversion of the grayscale iris image based on the max and min intensities of the actual iris pixels binary conversion of a grayscale iris image. The binary conversion is based on the minimum intensity of the pixels of the grayscale image.

[0049] Figure 13 illustrates an initial pupil binary coded image after minimum thresholding.

[0050] Figure 14 illustrates the global binary conversion of a grayscale image of an eye. [0051] Figure 15 illustrates the morphological opening of the global binary conversion image.

[0052] Figure 16 illustrated the binary converted pupil image after being image multiplied by the eyelash detected binary image.

DETAILED DESCRIPTION OF THE INVENTION [0053] While preferred embodiments of the present invention have been shown and described herein, it will be obvious to those skilled in the art that such embodiments are provided by way of example only. Numerous variations, changes, and substitutions will now occur to those skilled in the art without departing from the invention. It should be understood that various alternatives to the embodiments of the invention described herein may be employed in practicing the invention. It is intended that the following claims define the scope of the invention and that methods and structures within the scope of these claims and their equivalents be covered thereby. [0054] In general, the present invention provides systems and methods for rapid and robust identification of an individual based on the iris of the human eye. In various embodiments, a more rapid and robust identification is facilitated by pre-processing an iris image, whereby pre-processing includes but is not limited to removal of reflections, filling in holes in the image with average pixel intensities from the surrounding pixel intensities, filling in the holes in the image with a predetermined pixel intensity or filling in the holes in the image with a average intensity for the iris segmentation.

[0055] Any of the steps herein can be performed using computer program product that comprises a computer executable logic recorded on a computer readable medium. The computer program can work in any computer that may be any of a variety of types of general-purpose computers such as a personal computer, network server, workstation, or other computer platform now or later developed. In some embodiments, a computer program product is described comprising a computer usable medium having the computer executable logic (computer software program, including program code) stored therein. The computer executable logic can be executed by a processor, causing the processor to perform functions described herein. In other embodiments, some functions are implemented primarily in hardware using, for example, a hardware state machine. Implementation of the hardware state machine so as to perform the functions described herein will be apparent to those skilled in the relevant arts. Therefore, in various embodiments, systems of the invention comprise a computer executable logic on a computer readable medium which instructs certain functions to be performed involving both hardware/software operably linked to components in the systems. As used herein, the term "operably linked" means direct or indirect connections between two or more components of the system (e.g., computer central processor unit with both an input and output means). In various embodiments, a component of the methods or systems of the invention is a device is used to obtain an image, and the device is operably linked (e.g., LAN line, wireless, GPS, hardwire to computer hardware/software). For example, a digital image device is any camera, cellular telephone, personal digital assistance (PDA), video camera, camcorder, computer or other device having an optical sensor for digitally capturing an image. The optical sensor is any device that converts received light signals into digital signals representing the image that produces or reflects the light signals. In the exemplary embodiment, the optical sensor is a Charge-Coupled Device (CCD) or Complementary Metal Oxide Semiconductor (CMOS). The processor is any processor, microprocessor, computer, microcomputer, processor arrangement, or application specific integrated circuit (ASIC) suitable for performing the functions described herein. In most circumstances, a processor facilitates the functions and general operations of the digital image device in addition to performing the image processing functions.

[0056] For example, a conversion processor can evaluate the color correction metric which is based on a relationship between the standardized color space and the uniform human perceptual visual color space (CIELAB) after a color correction matrix is applied to the white balanced digital reference signal. The total cost function of the color correction metric is minimized to optimize the color correction and determine to optimum color correction parameters. [0057] In some embodiments, prior to conversion of a test image to grayscale, a test image can be processed for the appropriate color correction matrix to the white balanced digital reference signal results in a color corrected, white balanced, digital reference signal within a nonstandard RGB color space. By applying this signal to the appropriate gamma curve function, the signal is translated to an industrial standardized reference color space such as sRGB. To determine appropriate gamma curve parameters for the gamma curve function, a gamma curve optimization procedure is performed. [0058] The color, noise and contrast portions may be weighted to change the contribution, of each portion to the total cost function. In the exemplary embodiment the weighting is based at least partially on the particular gamma curve function. Therefore, a set of weighting values may be associated with each type of gamma curve function. [0059] A basic input/output system (BIOS), which contains the basic routines that help to transfer information between elements within the computer, is stored in the ROM. The computer also may include a hard disk drive for reading from and writing to a hard disk (not shown), a magnetic disk drive for reading from or writing to a removable magnetic disk, and an optical disk drive for reading from or writing to a removable optical disk, such as a CD ROM or other optical media. The hard disk drive, magnetic disk drive, and optical disk drive are connected to the system bus by a hard disk drive interface, a magnetic disk drive interface, and an optical disk drive interface, respectively. These drives and their associated computer-readable media provide nonvolatile storage of computer- readable instructions, data structures, program modules, and other data for the personal computer. It will be appreciated by those skilled in the art that other types of computer-readable media that can store data that is accessible by a computer, such as magnetic cassettes, flash memory cards, digital video disks, Bernoulli cartridges, random access memories (EAMs), read only memories (ROMs), and the like, may also be used in the example operating environment. [0060] In various embodiments, computer executable logic for conducting functions disclosed herein (e.g., image enhancement, iris/pupil segmentation, morphological opening, hole filling, reflection removal) are comprised on a medium operably linked to input, output and memory components which are conventional in the art. For example, a number of program modules comprising iris recognition methods described herein can be stored on the hard disk drive, magnetic disk, optical disk, ROM, or RAM, including an operating system, one or more application programs, other program modules, and program data. A user can enter commands and information into the computer through input devices, such as a keyboard 101 and pointing device (such as a mouse). These and other input devices often are connected to the processing unit through a serial port interface, for example, that is coupled to the system bus, but they also may be connected by other interfaces, such as a parallel port, game port, or a universal serial bus (USB), and the like. Further still, these devices maybe coupled directly to the system bus via an appropriate interface.

[0061] In some embodiments, computer executable logic functions to perform the various preprocessing, computations, conversion and analysis of test and reference iris images automatically, with any final or intermediate results capable of output using various conventional output devices. In some embodiments, an operator can manually perform functions described herein for iris recognition using computer executable logic to perform various functions as desired by the operator. A typical output device is a monitor or other type of display device also may be connected to the system. In some example environments, a stylus digitizer and accompanying stylus are provided in order to digitally capture freehand input. For example, the digitizer may be directly coupled to the processing unit, or it may be coupled to the processing unit in any suitable manner, such as via a parallel port or another interface and the system bus as is known in the art. Furthermore, the usable input area of the digitizer may be co-extensive with the display area of the monitor. Further still, the digitizer may be integrated in the monitor, or it may exist as a separate device overlaying or otherwise appended to the monitor.

{0062] The computer, comprising the necessary computer executable logic for performing functions described herein, can operate in a networked environment using logical connections to one or more remote computers, such as a remote computer . The remote computer can be a server, a router, a network PC, a peer device or other common network node, and it typically includes many or all of the elements described above relative to the computer. The logical connections include a local area network (LAN) and a wide area network (WAN)- Such networking environments are commonplace in offices, enterprise-wide computer networks, intranets, and the Internet, using both wired and wireless connections. [0063] In various embodiments, the input scan of an Ms is obtained in a remote location, transmitted to through a server to a system or subsystem (including a database), which comprises the reference image(s) against which the test image is compared using the computer executable logic which first functions to preprocess and enhance the test image as described herein. Furthermore, a computer comprising the image processing and enhancement software can be directly linked to the server, system, subsystem, such as through LAN, WAN, wireless, or hybrid land/satellite signaling. When used in a LAN networking environment, the computer is connected to the local area network through a network interface or adapter. When used in a WAN networking environment, the computer typically includes a modem or other means for establishing a communications link over the wide area network such as the Internet. The modem, which may be internal or external to the computer, may be connected to the system bus via the serial port interface. In a networked environment, program modules depicted relative to the personal computer, or portions thereof, may be stored in the remote memory storage device. [0064] Of course, it should be understood that one or more computers can be networked, wherein various embodiments of the invention are effected on the various computers on the network. It will be appreciated that the network connections and other techniques for establishing a communications link between computers can be used. The existence of any of various well-known protocols such as TCP/IP, Ethernet, FTP, HTTP, UDP, and the like is presumed, and the computer can be operated in a user-server configuration to permit a user to retrieve web pages from a web-based server. Any of various conventional web browsers can be used to display and manipulate data on web pages

[0065) Therefore, in various embodiments of the invention systems and methods of the invention comprise the necessary conventional components necessary to input/output information, convert, compute, compare., process and analyze data, utilizing the computer executable logic to effect iris recognition as described herein.

[0066) In one aspect of the invention, rapid and robust image processing facilitates large-scale iris recognition. In one embodiment, rapid and robust image processing is utilized to more accurately perform person identification on poor quality or non-ideal iris imagery.

[0067] In one aspect of the invention, image analysis algorithms preprocess, locate, and extract the iris structure from a digital image of the human eye. Furthermore, preprocessing may include iris and pupil segmentation. [0068] In various embodiments, reflections associated with the eye image are detected and removed in order to enhance segmentation process. The iris is extracted through a process of dynamic thresholding and edge detecting in order to generate a binary segmented iris. The dynamic threshold is set on the actual appearance of the iris tissue. For example, iris features are decomposed into a feature matrix through the process of a geometric transformation and structural enhancement. Pattern recognition is performed by computing the distance of two feature matrices of a binary iris template using the hamming distance metric formula. The distance metric measure positively establish, confirm, or disconfirm, the identity of any individual.

{0069] In one embodiment, iris and pupil segmentation comprises producing a grayscale iris and pupil image which is converted into binary images before boundary detection. In other embodiments, either the iris or pupil grayscale image is converted into binary images before boundary detection. In yet further embodiments, boundary detection for any iris or pupil segmentation is conducted prior to conversion from grayscale images to binary images (e.g., either for iris and pupil; or iris or pupil).

[0070] A recognition algorithm for rapid and robust identification of an individual based on the iris of the human eye. Iris technology has the smallest outlier (those who cannot use/enroll) group of all biometric technologies. Iris recognition provides a biometric authentication technology designed for use in a one-to many search environment, a key advantage of iris recognition is its stability, or template longevity as, barring trauma, a single enrollment can last a lifetime.

[0071] In one embodiment, image analysis algorithms preprocess, locate, and extract the iris structure from a digital image of the human eye. Reflections associated with the eye image are detected and removed in order to enhance segmentation process. The iris is extracted through a process of dynamic thresholding and edge detecting in order to generate a binary segmented iris. The dynamic threshold is set on the actual appearance of the iris tissue. Iris features are decomposed into a feature matrix through the process of a geometric transformation and structural enhancement. Pattern recognition is performed by computing the distance of two feature vectors of a binary iris template using the hamming distance metric formula. The distance metric measures positively establish, confirm, or disconfirm, the identity of any individual.

[0072] One aspect of the present invention is directed to a system or methods comprising a preprocessing to provide iris recognition more accurately and more rapidly. In one embodiment, preprocessing comprises reflection removal from a subject iris image prior to determining if the image matches a reference image (e.g., as contained in a database of images used to identify an individual's identity). Reflection removal removes any spurious high intensity reflections (e.g., corresponding to the illumination source) from the components of the eye in an iris image. Reflection removal is from cornea, sclera, and/or iris. [0073] In one embodiment, the reflection removal further comprises computing the global binary threshold that minimizes the intraclass variance defined by the following equation: [0074] σ(t) = qi(t)σ,(t) + q₂(t)σ₂(t)

[0075] Where σ(t) is the sum of weighted variances of the two modes of the histogram as a function of the threshold t. Further, σ(t) and q(t) be the respective variance and probability of one of the two modes of the histogram separated by a threshold t.

[0076] In one embodiment, the reflection removal further comprises computing the global binary threshold, an optimum threshold of an image that is used to create a binary (black/white) image through detecting and separating background pixels from foreground pixels. The image histogram is assumed to be bimodal, meaning it can be separated into two classes (background and foreground pixels). The normalized image histogram is treated as a discrete probability density function, as in

10077] p,<r_q) = n/n q-0, 1 , 2, ..., L-I

[0078] where n is the total number of pixels in the image, n_q is the number of pixels that have intensity level r_ς, and L is the total number of possible intensity levels in the grayscale image. The method then chooses the threshold value that maximizes the between-class variance O²B, which is defined as [0079] CT² _B = ω_o(μo - μt)² + ωi(μi - μ_τ)²

[0080] where k-1

G)₀ = ∑ Pq(^rq) q = 0

L-1 ω. = ∑ Pq<^rq) q = k k-1 Mo = ∑ qPq(r_q)/ ω₀ q= o

L-1 μ. = Σ qpq(r_qy ^ωi

q = o [0081] The threshold t that results in the minimization of σ(t) is used to convert the image to binary, and the threshold t that results in the maximization of O² _B is used to convert the image to binary (e.g., FIG. 14). [0082] Once converted to binary the morphological opening of thresholded image is computed using a circular(other elements could be used) structuring element as in: [_0083] A o B = (A θ B) ® B

[0084] Where A and B are the binary iris image and the circular structuring element respectively. The moφhological opening is simply an erosion (θ) of A by B, followed by the dilation (®) of the result by B. This morphological operation results in the opening of all large binary holes in the image and retains smaller holes left by the high intensity reflections and other artifacts (e.g., FIG. 15). [0085] The holes are then located in the binary opened image. If no holes are detected portions of the image that correspond to highest intensity regions are selected with the following pseudo code: [0086] high_intensity_threshold = max_intensity(grayscale_iris_image)

[0087] high_intensity_regions =Find(grayscale_iris_image <= (high_intensity_threshold - set_pixel_value)

[0088] The above pseudo code isolates the high intensity portions of the grayscale eye image based on a value that is a set intensity level lower than the highest intensity (bigh_intensity_threshold - set_pixel_value).

[0089] In various embodiments, such high intensity regions may be filled-in with an approximate value from the surrounding pixels, from an approximate average or median value for the entire iris.

[0090] In some embodiments preprocessing comprises segmentation of the pupil and/or iris, or any other portions of the eye obtained in an image. For example, in pupil segmentation the pupil is segmented from the iris at the pupillary boundary of an iris image. Furthermore, a grayscale image may be converted to binary for further analysis.

[0091] For example, for pupil detection the minimum intensity of a grayscale iris image is computed and converted to a binary based image on the minimum intensity detected (e.g., FIG. 13). Furthermore, an error check can be conducted for pupil detection (i.e., Pupil Detection Error Check), where for example, the minimum intensity is dynamically set by denoting a predefined amount of pixels that must be present, under the predefined minimum, in order to locate a binary pupil object. [0092] As described in the following pseudo code.

[0093] min_intensities==((Find (eye_image_pixel_intensities)) = (minimum_image_intensity))

[0094] While length(mm_mtensities)<mintmum_defined_iris_intensities [0095] {niinimum_iniage_mtensity= minimum_image_intensity+l min_intensities=((Find

(eye_image_pixel_intensities)) < = (minimum_image_intensity)) }

[0096] As the pseudo demonstrates, computer executable logic is designed to detect a minimal number of pixels that constitute at least a portion of the pupil pixels in the eye image. Each iteration detects more and more of the image pixels, however since the minimum is usually a good determination of the pupil few iterations are needed to detect the pupil.

[0097] In another embodiment, preprocessing comprises eyelash detection where the edges of the grayscale iris image are detected using the local gradient g(x,y) = [G² _* + G² _y]^1/2 and edge direction α(x,y) = tan^'1 (G_y/ G_x) computed at every point. The edge point is defined as a point whose strength is locally maximum in the direction of the gradient (e.g., FIG. 10) and the size of the edges is morphologically dilated with a circular structuring element to detect eyelashes (e.g., FIG. 11). Dilation is shown in the following equation:

_[0098] A θ B = {z\(B)_z n A ≠ 0} [0099] Where B is the structuring element and Λ is the binary edge image.

[00100] In another embodiment, preprocessing of a subject image comprises removal of spurious objects (to produce holes) and filling- in the holes using approximate values for surrounding regions or a median/average intensity value for the region in which such holes are produced. For example, removal of spurious objects comprises producing a morphological opening, Binary Filtering the image, and Binary Hole Filling.

[00101] In yet another embodiment, preprocessing comprises pupil selection where the largest circular object in a subject binary image is selected.

[00102] In one embodiment, preprocessing comprises eyelash removal. For example, eyelash removal comprises binary multiply with eyelash detected binary image (Fig. 11) and binary pupil image (Fig. 13): [00103] eyelash_removed_binary_image = image_multiply(binary_pupil_thresholded_image, eyelash_detected_binary_image)

[00104] The result of the image multiplication removes eyelashes present in the image binary pupil image (FIG. 16). [00105] In some embodiments, a method of the invention for iris recognition comprises pupil segmentation. For example, pupil segmentation comprises morphological closing and hole filling to complete an iris structure. Furthermore, pupil segmentation can comprise detecting the largest circular object boundary (segmentation of the pupil) and computing the center coordinates of pupil and pupil's horizontal radius at its center. [00106] In another embodiment of the invention, a method of the invention for iris recognition comprises performing iris segmentation. For example, Iris —the iris is segmented from the sclera at the limbic boundary of an iris image. This may be followed by iris pixel intensity selection.

[00107] In addition, iris segmentation may comprise dynamically defining regions of interest (ROI) outside of the east and west side boundaries of the pupil segmentation; converting the 2-Dimensional ROI into a mean intensity signal; computing the horizontal derivative of the regions of interest; locating the maximum peak (corresponds to an approximate limbic boundary); defining a localized ROI based on the approximate limbic boundary; and computing the max, mean, and min intensities within the localized ROI.

[00108] Furthermore, iris segmentation, may comprise performing morphological operations to remove reaming spurious features and hole; followed by boundary detection of the largest circular object (segmentation of the iris); and computing the center coordinates of iris and iris's horizontal radius at its center . [00109] In various embodiments of the invention disclosed herein, iris recognition comprises binary image conversion, such as through converting a Grayscale iris image to binary using the maximum and minimum intensities of the localized ROI.

[00110} In any of the methods/systems described herein , circular segmentation (e.g., iris, pupil) can comprise computing a circular boundary around the iris based on the limbic boundary approximation; and removal of all features outside of the boundary as in Fig.9. [00111] Another aspect of invention comprises performing normalization including but not limited to transformation, image enhancement and/or template mask generation. Thus in one embodiment, transformation comprises converting a circular iris segmentation to a rectangular orientation (Fig. 1 Iris structure normalization), followed by computing the iris radius (e.g., distance between the limbic and pupillary boundaries of the iris image), and computing the geometric transformation (e.g. , unwrap the iris from a circle to a rectangle) based on the center coordinates of the iris and pupil, and iris radius. [00112] In another embodiment, normalization comprises image enhancement as shown in Fig. 1. Iris Structure Enhancement. For example, enhancement of the transformed iris's textural content can comprise filtering the unwrapped image with an averaging filter, subtracting the filtered image from the original unwrapped image to enhance structural content; and minimum-maximum normalization to normalize the intensities of the image. S [00113] In one embodiment, normalization comprise performing a template mask generation (Fig. 3 Iris

Structure Mask). For example, a mask is computed based on the segmentation that blocks all unwanted features that do not pertain to the iris.

[00114] It should be evident that in various embodiments, normalization can comprise one or more embodiments disclosed above, such as transformation, image enhancement and/or template mask generation. In 0 certain instances the image obtained may be of poor quality (e.g., imaging device/poor lighting, etc.). However, through practice of the normalization methods described herein, even poor quality images may result in positive match as illustrated in FIGs. 5 and 6. These graphs indicate that the derived quality metric substantially reduces or eliminates matching errors that are due to very poor quality iris images in the recognition system. [00115] In various embodiments, a predetermined quality metric is set so that a prescribed data set of images S (e g-, test image and data set of references images) is output test image. For example, if a high security application is required, the data set can be predetermined by size (e.g., output closes matches based on the preset quality metric). In this way if the quality of the test image(s) is poor, the top matches are returned in the output for an operator to analyze. In other words, computer executable logic of the invention can function to select a prescribed data set for output by a preset quality metric. 0 [00116] Another aspect of the methods of the invention for iris recognition comprise pattern recognition. For example, pattern recognition comprises encoding- Wavelet filtering of an image in order to extract structural features into a binary template. Furthermore, pattern recognition may comprise computing a feature metric- hamming distance based on the normalized iris image template and template mask. Thus, a determination is made as to a positive or negative match where there is a score greater than a determined threshold it indicates a match and5 if it is less than a threshold then it indicates failure (i.e., no match).

[00117] The threshold can be set as desired for a particular application. For example, for a heightened security setting (e.g., entrance to a military facility, versus access to a home computer) the threshold can be set to provide a trade off between the false match rate (also known as the false accept rate) and the false reject rate (FIG.7 provides an example of probability distributions). For example, a test data set is acquired with a specific camera and the0 match score that falls slightly less than the Equal Error Rate (EER) is selected, which is the point where the false accept rate is equal to the false reject rate and where there should be no false matches. Furthermore, the matching threshold can be set drastically lower than the EER to where it would be virtually impossible to get a false match. Such a low threshold would generally be used in high security applications, because the false reject rate (meaning that a lot of true matches are denied simply due to the threshold setting) will be very high. For example, if iris5 recognition is being used for a relatively lower security application (e.g., access to a computer) it may be preferred to obtain verification on the first try as opposed to getting rejected, for example, 5 times before verification. The methods/systems of the present invention are adaptable so that a desired matching threshold can be set. Performance can be quality tested for any given threshold (FIGs. S and 6) to determine where the false match rate (FMR) equals the false reject rate (FRR) for the threshold. [00118] In one embodiment, an iris recognition method is provided for uniquely identifying a particular human being by biometric analysis of the iris of the eye, comprising the following steps: a. preprocessing an acquired image of an eye of said human to be identified; b. obtaining pupil segmentation information; c. obtaining iris segmentation information; d. normalizing said pupil segmentation and iris segmentation information to obtain a pattern recognition code; and e. quality assessment of said normalization, iris segmentation, pupil segmentation, preprocessing, and acquired image of an eye to assess feasibility of matching f. comparing said pattern recognition code with a reference iris code to determine to uniquely identify said particular human being if said pattern recognition and reference iris code match. [00119] In further embodiments, preprocessing comprises at least removing reflection corresponding to an illumination source from said acquired image; or computing a global binary threshold corresponding to said acquired image in order to convert an intensity image of the eye into a binary coded image.

[00120] In further embodiments, a method of iris recognition comprises a step of producing a morphological opening of said threshold to obtain an opened image that isolates the reflection from the noise in the image; and further comprising identifying one or more holes in said opened image (e.g., corresponding to highest intensity

(reflection) regions); or selecting said one or more holes and filling said holes with an approximate value from surrounding pixels.

[00121] In various embodiments of the methods of the invention for iris recognition, performing pupil segmentation comprises segmenting said pupil from said iris at the pupillary boundary. Furthermore, an iris image is provided in grayscale or converted to grayscale.

[00122] For example grayscale images can be manipulated further, such as in computing a minimum intensity of said grayscale iris image and converting said image to binary code based on said minimum intensity.

[00123] In one embodiment, a method of the invention comprise performing a pupil detection error check, such as through computing Euclidean Distance Transform (EDT) from said grayscale image to the nearest nonzero pixel and converting said iris image to binary code using said EDT.

[00124] In an addition, the method can further comprising removing spurious objects and filling said holes.

For example, filling comprises binary hole filling. In addition removing may comprise binary filtering.

[00125] In addition, methods of the invention may comprise performing one or more functions as a quality metric. For example, in one embodiment, a quality metric comprises eyelash detection to improve accuracy. Eyelash detection can comprise detecting edges of said grayscale iris image; or wherein said eyelash detection comprises filtering and dilating said edges to detect said eyelash. As such, interference from eyelash structures is recognized and reduced or eliminated.

[00126] In other embodiments, a method of iris recognition is provided comprising the step of pupil selection.

For example, pupil selection comprises selecting the largest circular binary object in said image. If a pupil is not detected, a minimum intensity is scaled up until a binary object is detected, e.g., corresponding to a certain pupil area.

[00127] In one embodiment, methods of the invention further comprise removing eyelash structures, e.g., by performing binary multiplication with eyelash detected binary image.

[00128] In any of the methods disclosed herein, pupil segmentation may comprise morphological closing and hole filling to complete a pupil structure and detecting the largest circular object Furthermore, such a method may comprise measuring center coordinates and horizontal radius corresponding to said pupil. [00129] In any of the iris recognition methods disclosed herein, an iris segmentation comprises segmenting said iris from the sclera at the limbic boundary of said iris image. In addition, iris segmentation may comprise selecting iris pixel intensity. Furthermore, selecting iris pixel intensity comprises defining regions of interest outside of east and west boundaries of said pupil segmentation. Alternatively, iris recognition may comprise converting said region of interest into a mean intensity signal and measuring the horizontal sampled derivative of said regions of interest.

[00130] Furthermore, such a method can comprise identifying a maximum peak corresponding to an approximate limbic boundary, defining a localized region of interest based on an approximate limbic boundary and computing a maximum, mean and minimum intensities within said localized region of interest. [00131] In one embodiment, maximum and minimum intensities are utilized to obtain a binary of said grayscale iris image.

[00132] In various embodiments, preprocessing an image comprises performing circular segmentation, such as measuring a circular boundary around said iris based on a limbic boundary approximation and removing one or more features outside said boundary. [00133] In practicing a method of the invention, it may be desirable to further remove reaming spurious features and holes in said iris image (e.g., in performing pupil or iris segmentation). For example, in such a method iris segmentation comprises detecting a largest circular object boundary; and measuring center coordinates and horizontal radius corresponding to said iris. [00134] In one embodiment, a method for iris recognition is provided, comprising the steps of acquiring an image of at least one eye of a user, and of processing said image to remove reflection, conversion of said image to grayscale to measure minimum and maximum intensities, determining the outer boundary of the iris from said minimum and maximum intensities, filling in holes with intensities from surrounding pixel intensities; and determining if the processed image matches to a reference image. [00135] In another embodiment, a method for iris recognition comprises: providing an image of an eye; selecting a pupil in the image; segmenting the pupil; selecting an iris in the image; segmenting the iris; wherein said segmenting detects and removes reflections in said image to enhance said segmentation; and determining if said image matches to a reference image.

[00136] In any of the embodiments disclosed herein, methods of iris recognition can comprise a step of obtaining a quality metric. Obtaining a quality metric includes but is not limited to determining image quality parameters of pupil segmentation, occlusion of the iris, size of the pupil dilation, pupil constriction, number of pixels inside the iris and clarity of iris pixels. For example, a quality metric is a score (0-100) calculated based on the following features of the algorithm and image: blurriness/noisiness of the image, pupil segmentation circularity, iris segmentation circularity, iris pixel resolution, and occlusion estimates. [00137] In addition, for methods disclosed herein, iris recognition can comprise iris or pupil (or iris and pupil) segmentation comprising converting a subject image into a binary coded image. Furthermore, such conversion may be performed before or after a step comprising determining a pupil or limbic boundary. In one embodiment, such conversion is performed after such determination.

Claims

WHAT IS CLAIMED IS:

1. An iris recognition method for uniquely identifying a particular human being by biometric analysis of the iris of the eye, comprising the following steps: a. preprocessing an acquired image of an eye of said human to be identified;

S b. obtaining pupil segmentation information; c. obtaining iris segmentation information; d. normalizing said pupil segmentation and iris segmentation information to obtain a pattern recognition code; e. quality assessment of said normalization, iris segmentation, pupil segmentation, preprocessing,0 and acquired image of an eye to assess feasibility of matching; and f. comparing said pattern recognition code with a reference iris code to determine to uniquely identify said particular human being if said pattern recognition and reference iris code match.

2. A method for iris recognition, comprising the steps of acquiring an image of at least one eye of a user, and of processing said image to remove reflection, conversion of said image to grayscale to measure minimum and5 maximum intensities, determining the outer boundary of the iris from said minimum and maximum intensities, filling in holes with intensities from surrounding pixel intensities; and determining if the processed image matches to a reference image.

3. A method for iris recognition comprising: providing an image of an eye; selecting a pupil in the image; segmenting the pupil; selecting an iris in the image; segmenting the iris; wherein said segmenting detects and removes reflections in said image to enhance said segmentation; and determining if said image matches to a reference image.

4. A computer product comprising computer executable logic for configured for iris recognition, wherein said logic processes a test eye image for iris and pupil segmentation, removes reflection(s) in said image to produce an enhanced image; compares said enhanced image to a reference image to determine if there is a match.