Adaptive iris matching using database indexing

Abstract

An adaptive iris matching approach for processing images of irises having a quality not sufficient for conventional non-adaptive matching approaches. Visible regions in a radial direction on an iris, without segmenting a circumferential of the iris, may be processed. A boundary of the visible region of the iris may be estimated. An iris map may be constructed with the non-visible portion of the iris masked. The iris map may be at least partially encoded. Partial codes of the iris map may be extracted to index at least portions of a database containing iris information. Irises may be retrieved from the database with an iris code as a query.

Description

BACKGROUND

The present invention pertains to recognition systems and particularly to biometric recognition systems; in particular the invention pertains to iris recognition systems.

SUMMARY

The invention is an adaptive iris matching approach for processing images of irises having a quality not sufficient for conventional non-adaptive matching approaches. Visible regions in a radial direction on an iris, without segmenting a circumferential of the iris, may be processed. A boundary of a visible region of the iris may be estimated. An iris map may be constructed with the non-visible portion of the iris masked. The iris map may be partially encoded. Partial codes of the iris map may be extracted to index at least portions of a database containing iris information. Irises may be retrieved from the database with an iris code as a query.

BRIEF DESCRIPTION OF THE DRAWING

FIG. 1 a is a diagram of an iris showing various delineated boundaries;

FIG. 1 b is a diagram of a blurred iris image;

FIG. 1 c is a diagram of a gazed eye;

FIG. 2 a is a diagram of polar segmentation of an iris in an image;

FIG. 2 b is a diagram of processing blocks for an adaptive iris matching approach;

FIG. 2 c is a diagram of adaptive matching;

FIG. 3 a is a diagram of polar segmentation used to estimate a pupil bound;

FIG. 3 b is a diagram of model fitting using a pupil warp function to obtain a symmetric pupil shape and outside iris boundary;

FIG. 3 c is a diagram of an inverse projection or transform as applied to iris, pupil and eyelid boundaries; and

FIG. 4 is a diagram of an operation of an iris image processing approach.

DESCRIPTION

While much progress has been made toward recognition through iris analysis under small variations in pose and lighting, reliable techniques for recognition under more extreme non-cooperative environments appear to have proven elusive. Iris recognition signatures may perform well when the data acquisition parameters are relatively constrained and the acquired data quality is moderately high. However, when the data acquisition and quality constraints are relaxed, the match performance of iris recognition may suffer a considerable decline.

This may be due in part to the variability of the resulting non-ideal iris relative to the constrained signature, and the likelihood that there is insufficient iris exposure to adequately normalize—a procedure required by virtually all existing iris algorithms. Apart from estimating these irregularities, the segmentation routine should also detect reflections due to ambient light sources and occlusions due to eyelashes and eyelids. The challenges of processing such irises may mount mostly in the segmentation process or a loss of iris information due to gazing or heavy obscurations.

One may introduce an adaptive solution based on partial segmentation and database indexing. A key aspect of the present approach may be that it does not necessarily require the usual upfront normalization procedure in order to capture these local variations and thus there is no need to estimate the entire outer border of the iris prior to matching. It is important to note that such a procedure may be crucial to virtually all existing iris recognition techniques.

One may introduce a technical approach to solve the above challenge by initially parsing through the portion of the iris that is visible, and then adaptively extracting features from the rest of the iris based upon data indexing and model fitting approximation. Present solution may extend the POSE technique (see e.g., U.S. patent application Ser. No. 11/043,366, filed Jan. 26, 2005, and entitled “Iris Recognition System and Method”) having capabilities to adaptively extract features from the iris without the prior need to estimate the actual outer bound of the iris. It is arguable that it may suffice only to capture local sharp variations along the radial directions in the nonobscured areas of the iris. The key aspect of the present approach is that it does not require a normalization procedure in order to capture these local sharp variations. It is important to note that such a procedure may be crucial to virtually all existing iris recognition techniques. One may introduce an approach of an iris recognition technique that does not necessarily require normalization of the iris map.

The approach is based on the fact that local details of the iris are spread along the radial directions and thus virtually any visible iris region may be independently processed on the radial direction without estimating the circumferential of the iris.

The merit of the present approach is that it may allow users to enroll or match poor quality iris images that would be rejected by other states of the art technology. In addition, this approach may generate a list of possible matches instead of only the best match when the iris-print is too obscured to be a reliable matcher. Another merit of the present approach is that it may allow the iris recognition system to be more tolerable of noise and could potentially improve the matching process computational efficiency. The database may be easily indexed using a skin-print or eye-print to fewer manageable subsets of individuals.

One may solve the problem in three processing steps. 1) One may first segment only the visible region. Most of the fitting models may provide a reasonable estimate of the border segment to a first order estimation. 2) One then may extract the associated partial codes of the iris barcode to index a smaller population in a large population. 3) One then may recover the noisy information in the obscured area by adapting the appropriate fitting model that provides the best matching bits within the limited subset retrieved in step 1). The last processing step may be beneficial to denoise to some extent the obscured area as long as the iris information is not lost due to heavy gazing or obscurations.

In step 3, alternatively, one may capture the iris dyadic variations by formulating a dynamic encoder that takes into account the varying edge points. Due to the dyadic nature of wavelet functions, one may modify the encoding scheme to make use of a dynamic wavelet function that varies as a function of the angular variable. Unlike the standard wavelet transform, the wavelength scalar is not necessarily a constant but rather may be a smooth continuous function as a function of the angle variable. The output of the wavelet filter may thus be evaluated at each angle using virtually all possible values for the best match. The concatenated matched bits may determine the overall match score between a probe and query barcode.

The initialization step for the adaptive iris model may be based on the perspective projection of an ideal frontal pupil image (i.e., circular shape) into the actual image plane. Since the pupil edges are visible and easy to estimate, one may compute the warping projection function. One may use the inverse transform of the projection on both the pupil and iris edges to estimate the iris boundaries.

One may introduce a technical approach to solve the above problem by initially parsing through a portion of the iris that is visible, and then adaptively extracting features from the rest of the iris based upon data indexing and model fitting approximation. The solution may extend the POSE technique capabilities to adaptively extract features from the iris without the prior need to estimate the actual outer bound of the iris. Arguably, it may suffice only to capture local sharp variations along the radial directions in the non-obscured areas of the iris. A key aspect of the present approach is that it does not necessarily require a normalization procedure in order to capture the local sharp variations. It is important to note that such a procedure is crucial to virtually all existing iris recognition techniques. To one's best knowledge, one may introduce the first approach of an iris recognition technique that does not require normalization of the iris map.

A merit of the present approach is that it may allow users to enroll or match poor quality iris images that would be rejected by the current state of the art technology. In addition, this approach may generate a list of possible matches instead of only the best match when iris-print is too obscured to be a reliable matcher. Another merit of the present approach is that it may allow the iris recognition system to be more tolerable of noise and could potentially improve the matching process computational efficiency. The database may be easily indexed using skin-print or eye-print to fewer manageable subsets of individuals.

The original foundation of the iris technology and algorithms do not appear to address issues such as side looking eyes, self occlusion, non-frontal face images, and so forth. Recognition of this situation has led to solutions to address the real-time operational requirements of a standoff iris recognition system. The present approach is an efficient and robust one-dimensional (1D) fast segmentation approach built around the present approach for the Polar Segmentation (POSE) system (see e.g., U.S. patent application Ser. No. 11/043,366, filed Jan. 26, 2005, and entitled “Iris Recognition System and Method”).

There may be great promise for iris technology with a false reject rate between 0.01-0.03 at a false acceptance rate of 10-3. The false non-matching rate (FNMR) versus false matching rate (FMR) may be better even on larger datasets. The uniqueness and richness of iris patterns when deployed to match billions of subjects may be a requirement that should be met to deploy the technology for things such as border control and the global war on terror.

A challenge to solve may be presented. Current iris recognition solutions do not necessarily reflect the full potential of the technology. Iris recognition signatures may perform well when the data acquisition parameters are relatively constrained and the acquired data quality is moderately high. However, when the data acquisition and quality constraints are relaxed, the match performance of iris recognition may suffer considerably. Very few developed techniques appear to address the true nature of iris irregularity in iris segmentation. This may be due in part to the variability of the iris-print in unconstrained scenarios and less than optimal acquisition conditions, as illustrated by the examples shown in FIGS. 1a, 1b and 1c. Apart from estimating these irregularities, the segmentation routine may also detect reflections due to ambient light sources and occlusions due to eyelashes and eyelids. The difficulties of processing such irises may reside mostly in the segmentation process, loss of spatial frequency content (e.g., blur side effect), or loss of iris information due to gazing or heavy obscurations.

As shown in FIGS. 1a, 1b and 1c, the pupil edges and the eyelids (highlighted at items 11 and 12) may be estimated. The iris borders, on the other hand, may yield multiple fitting models 1, 2 and 3, as noted with numerals 13, 14 and 15, respectively. The uncertainty in the iris segmentation may result in assigning bit codes to the wrong iris pixels during feature extraction, and therefore drastically affecting the matching outcome. Similarly, the blurry low-quality image 16 in FIG. 1b appears to have no high spatial frequency content of the actual iris pattern. The strong gazed eye 17 in FIG. 1c appears to have led to a compression of the iris information that cannot be recovered. An image acceptable to current commercial iris technology may generally require highly constrained conditions such as little to no gaze, high contrast, and rich spatial frequency content. Thus, these illustrative examples may represent a challenge for iris recognition technologies.

FIGS. 1 a, 1b and 1c, from left to right, show examples of challenging cases of iris images: (1a) obscured eye 18 (1b) blurred eye 16 (boxes 19 and 21 indicate areas with key holistic ocular features), (1c) strong gazing 17 (patch(es) 22 indicate discriminating skin textures). No existing iris approach appears to address these types of problems simultaneously, as many seem to rely solely on visible iris-prints and constrained environments.

The present approach is an iris recognition system implementing image quality metrics as a first step of the process to assess the quality of the acquired eye image for reliable operation, and then apply the appropriate iris processing. Images with high quality may be processed using the POSE technique or virtually any third party reliable iris recognition system.

If the iris image does meet a subset of quality metrics (e.g., not blur but obscured images), then the eye image may be passed to the adaptive iris matching process described below. If the iris image does not meet a second set of quality metrics and cannot be rehabilitated, then the eye image may be rejected and a new eye image capture be sought by the eye-finder.

The present technical approach for iris quality measurement is described in U.S. patent application Ser. No. 11/681,614, filed Mar. 2, 2007, and entitled “Iris Recognition System Having Image Quality Metrics”.

An adaptive iris matching technique (AIM) may be noted. Very few techniques appear to have been developed to address the true nature of iris irregularity in standoff iris segmentation. Apart from estimating these irregularities, the segmentation routine should also detect reflections due to ambient light sources and occlusions due to eyelashes, eyelids and gazing.

The present approach may be based on the fact that local details of the iris are spread along the radial directions, and thus virtually any visible iris region may be independently processed on the radial direction without segmenting the circumferential of the iris. It may suffice only to capture local variations along the radial directions in the non-obscured areas of the iris for inferring an iris biometrics. A key aspect of the present approach is that it does not necessarily require the usual upfront normalization procedure in order to capture these local variations; thus, there appears no need to estimate the entire outer border of the iris prior to matching. Such a procedure may be crucial to virtually all existing iris recognition techniques, and use the segmentation functionalities available in the POSE system. The present approach is illustrated in FIGS. 2a, 2b and 2c. FIG. 2a is a diagram of POSE segmentation. FIG. 2b shows AIM processing blocks partial encoding 25, database indexing 26 and adaptive coding 27. The blocks effectively show the following: (1) segment only visible iris regions; (2) index to a manageable database; and (3) search for a best matching model using adaptive encoding. FIG. 2c is a diagram of adaptive matching.

The present approach may solve the issue in three processing steps.

1) Partial encoding 25 may be noted. One may first segment only the visible region (e.g., the iris segment highlighted at 23 in FIG. 1a, as delineated by lid borders 12). The fitting model 15 in FIG. 1a may provide a reasonable first order estimation of the border segment. Since one may estimate the iris boundaries for the visible iris segment 23, one can construct the iris map using a “rubber sheet” model while masking the rest of the iris. One may then proceed to encode the phasor information of the extracted iris map using Log-Gabor or Wavelet filters so that comparisons between templates can be made at least partially.

2) Database indexing 26 may be noted. One may extract the associated partial codes of the iris barcode to index a smaller population in the large database. Verification of an iris barcode against large biometrics databases may be daunting, especially when one has only partial code for an iris. To retrieve irises from a large database, one may use the extracted iris code as a query and compute two feature vectors for the iris query (phasor based clustering and angular clustering). Organization of feature data may be completed hierarchically to compare one feature at a time, starting with the extracted feature elements that are closest to the inner iris boundary and going outward. Current linear database indexing does not necessarily leverage the data structure of the iris barcode. In one approach, one may use the KD-Tree search approach (e.g., see J. L. Bentley, “Multidimensional Binary Search Trees Used for Associative Searching,” Comm. of the ACM, p. 18(9), 1975). Nearly any other alternative search process may be applied to this process step. A primary benefit of the KD-Tree algorithm is that it may enable the use of spatial structure within the data to limit the cost of the search. In one approach, one may use structures from multiple features, and their indexing to increase the efficiency of the search.

Irises may be retrieved from a large collection of databases using a visible iris barcode as a query and then finding the similarity measure for all the features of the iris codes in the database. A search process may be noted in U.S. patent application Ser. No. 11/681,751, filed Mar. 2, 2007, and entitled “Indexing and Database Search System”.

3) Adaptive Encoding 27 may be noted. Appropriate segmentation of the rest of the iris may appear critical to the success of the iris recognition. Since data that is rescaled (e.g., replacing Model 1 (13) with Model 3 (15) in FIG. 1a) iris patterns may corrupt the iris code and the generated template, resulting in poor recognition rates, especially if the visible region is much smaller then the obscured regions. One may linearly adapt the fitting model that provides the best matching bits within the limited subset retrieved in Step 1 (25). The iris pattern deformation due to gazing could generate lateral pressure on some of the tissue that may cause small parts of the iris tissue to fold and change in patterns and some of which might be obscured by the eyelids. The approach may be to adaptively normalize the iris to adjust linearly the rubber sheet iris map to fit fewer identified models, as shown in FIG. 1a, and to align the map to the varying edge points of the model while searching for the minimum possible matching distance. The search may be executed only on the subset of subjects indexed at Step 2 (26) of the process. It is important to note that the map calibration may be applied not necessarily uniformly but rather linearly on the radial direction to fit the varying edge points of the matching model.

Alternatively, in another approach, in Step 3 (27) one may capture the iris dyadic variations by formulating a wavelet based encoder that takes into account the varying edge points. Due to the dyadic nature of wavelet functions, one may modify the encoding scheme to make use of a dynamic wavelet function that varies as a function of the angular variable,

$\begin{array}{cc}{W}_{\alpha }\left(R\right)={\left[\alpha \left(\theta \right)\right]}^{-1/2}\int \mathrm{sig}\left(r\right){\psi }^{*}\left(\frac{r-R}{\alpha \left(\theta \right)}\right)dr,& \left(1\right)\end{array}$ where R is the translation parameter and α is the scale parameter. The function ψ(·) may be the mother wavelet. Unlike the standard wavelet transform, the wavelength scale α(θ) may be a smooth continuous function of the angle. The concatenated matched bits may determine the overall match score between a probe and query barcode.

The initialization step for the adaptive iris model may be based on the perspective projection of an ideal frontal pupil image (i.e., circular shape) into the actual image plane. Since the pupil edges are visible and easy to estimate, one may compute the warping projection function. One may use the inverse transform of the projection on both the pupil and iris edges to estimate the iris boundaries as shown in FIGS. 3a, 3b and 3c. Polar segmentation may be used to estimate a pupil bound 31 in FIG. 3a. FIG. 3b shows a model fitting using pupil warp function to obtain a symmetric pupil shape 32 and outside iris boundary 33. The inverse projection or transform is applied to the iris 34, pupil 35 and eyelid 36 boundaries as shown in FIG. 3c. In the preprocessor approach, the iris signature is directly extracted from the reference warped image. By incorporating the warping projection in the adaptive iris matching algorithm, one may report an improvement of the matching accuracy at fixed equal error rate.

The following numbered items are information of the present system. 1) A standoff iris recognition system may have an iris image source, a quality processing module connected to the image source, where the quality processing module is for providing a quality assessment of an iris image.

2) In the system of item 1, the assessment of an iris image may be provided according to one or more of the following image conditions which include blur, obscuration, standoff angle, iris visibility, and so forth.

3) The system of item 2, may further have an evaluator connected to the quality measure, where the evaluator is for receiving a quality assessment of an iris image from the quality processing module, and for indicating whether the iris image is high quality, acceptable (can be processed using adaptive iris matching) or unacceptable for further processing, and the evaluator is for indicating whether an iris image that is unacceptable should be rejected or be rehabilitated for further processing, and the evaluator is for indicating whether an iris image that is acceptable should be processed using the adaptive iris matching process.

4) In the system of item 3, further processing may include segmentation of the iris image using adaptive process having partial encoding, database indexing, and adaptive matching.

5) The system of item 3, where the system medium that provides instructions of item 4 that, when executed by a machine, may further cause the machine to perform operations including normalizing a number of data points from the iris portion along the visible radial at an angle segment.

6) The recognition system that provides instructions of item 5 that, when executed by a machine, may further cause the machine to perform operations to encode the partial visible iris map signature.

7) The system of item 1 may have a approach of indexing: providing a database of templates; grouping the database into a plurality of sub-databases, where: each template in the database has a first number of bits (based on angular indexing), each sub-database of the plurality of sub-databases represents virtually all templates in the database, and each template in a sub-database has a second number of bits (hierarchical structure item, e.g., KD Tree approach).

8) The system of item 7 may further have providing a subset of possible template barcodes for matching, and selecting a second number of bits of a barcode that corresponds to the second number of bits of identified templates.

9) In the system of item 8, providing instructions of item 6 that, when executed by a machine, may further cause the machine to perform operations having normalizing a number of data points from the obscured iris portion to match one of the identified templates (Adaptive Encoding item: The best matching bits within the limited subset retrieved by the database indexing).

10) The recognition system that provides instructions of item 9 that, when executed by a machine, may further cause the machine to perform operations to encode the partial obscured iris map signature. (I.e., it may adaptively normalize the iris to adjust linearly the rubber sheet iris map to fit fewer identified models.)

11) The system of item 1 that provides instructions of item 10 that, when executed by a machine, may further cause the machine to perform operations having comparing the iris barcodes to a previously generated reference iris signature, indicated by item 8. (It may be an item on aligning the map to the varying edge points of the model while searching for the minimum possible matching distance).

The present standoff iris recognition approach using adaptive iris matching (AIM) may tackle the challenging case of iris recognition from images where irises are obscured but partially visible. One may provide an approach that may adaptively extract features from the iris without requiring the actual segmentation of the outer bound of the iris. It may suffice only to capture local sharp variations along the radial directions in the non-obscured areas of the iris. An aspect of the present approach may be that it does not require a normalization procedure in order to capture these local sharp variations. Such a procedure may be crucial to iris recognition techniques. One may introduce an approach of an iris recognition technique that does not require normalization of the iris map.

FIG. 4 is a diagram of an operation of the present approach. A quality evaluation at symbol 42 may be made of an iris image from an iris image acquisition module at symbol 41. If the quality of the image is deemed to be “high” (by reaching a predefined iris quality measure adequate for an iris to be processed using iris signature), the image may be processed using a non-adaptive iris matching approach at symbol 43. Such an approach may be POSE noted herein, or some commercial approach. If the quality of image is deemed to be “low” (does not reach a predefined iris quality level that allows recognition of an identity using iris signature), a second quality evaluation of the image may be made at symbol 44. A result of the second evaluation may be “acceptable low”, such as there is enough percentage of the iris region visible that can be processed, which indicates that the image may be processed by the present adaptive iris matching approach 58 starting at symbol 45. This image quality would not be acceptable to the non-adaptive iris matching approach at symbol 43. An alternative result of the second evaluation at symbol 44 may be “very low”. Here, the image may be rehabilitated at symbol 46. Because of something like a glitch or other issue with the image which could be cleared up, a result may be a “rehabilitated acceptable low” which means that the image may go to the adaptive iris matching approach at symbol 45. The result of rehabilitation at symbol 46 could be a “rehabilitated high” which means that the image could go to the non-adaptive iris matching approach at symbol 43. Or the image may be deemed as an “unrehabilitatable very low” which means that the iris image would be rejected at symbol 47.

At symbol 45, the approach for adaptive iris matching may begin. The following is a summary of the approach for illustrative purposes but could incorporate additional details. There may be a processing of virtually any visible iris region on a radial direction at symbol 48, without segmenting a circumferential of the iris of the image. The iris boundary may be estimated for the visible iris region at symbol 49. An iris map may be constructed using a rubber sheet model at symbol 51. Rest of the iris map may be masked. Phasor information of the iris map may be encoded for at least partial comparisons between templates, as indicated at symbol 52. Associated partial codes of the iris code may be extracted to index a portion of a database at symbol 53. Irises, i.e., images of them, may be retrieved from the database with the iris code as a query and determine a similarity measure for features of iris codes in the database at symbol 54. At symbol 55, the remainder of the iris may be segmented. The iris may be adaptively normalized to linearly adjust the rubber sheet iris map to fit fewer identified models (i.e., narrow the search) at symbol 56. The search may be executed just on the indexed portion of the database at symbol 57.

Relevant applications may include U.S. patent application Ser. No. 11/675,424, filed Feb. 15, 2007, and entitled “Standoff Iris Recognition System”; U.S. patent application Ser. No. 11/372,854, filed Mar. 10, 2006, and entitled “Invariant Radial Iris Segmentation”; U.S. patent application Ser. No. 11/043,366, filed Jan. 26, 2005, and entitled “Iris Recognition System and Method”; U.S. patent application Ser. No. 11/672,108, filed Feb. 7, 2007, and entitled “Approaches and Apparatus for Eye Detection in a Digital Image”; U.S. patent application Ser. No. 11/681,614, filed Mar. 2, 2007, and entitled “Iris Recognition System Having Image Quality Metrics”; U.S. patent application Ser. No. 11/681,751, filed Mar. 2, 2007, and entitled “Indexing and Database Search System”; U.S. patent application Ser. No. 11/681,662, filed Mar. 2, 2007, and entitled “Expedient Encoding System; U.S. patent application Ser. No. 10/979,129, filed Nov. 3, 2004, entitled “System and Method for Gate Access Control”; U.S. patent application Ser. No. 10/655,124, filed Sep. 5, 2003, and entitled “System and Method for Dynamic Stand-Off Biometric Verification” (issued as U.S. Pat. No. 7,183,895); and U.S. Provisional Patent Application 60/778,770, filed Mar. 3, 2006, and entitled “Stand-Off Iris Detection Tracking and Recognition System”; all of which are hereby incorporated by reference.

In the present specification, some of the matter may be of a hypothetical or prophetic nature although stated in another manner or tense.

Although the present system has been described with respect to at least one illustrative example, many variations and modifications will become apparent to those skilled in the art upon reading the specification. It is therefore the intention that the appended claims be interpreted as broadly as possible in view of the prior art to include all such variations and modifications.

Claims

1. A standoff iris recognition system comprising: an iris image acquisition module;an image quality evaluator connected to the iris image acquisition module; andan adaptive iris matching module connected to the image quality evaluator and the eye image acquisition module; andwherein the image quality evaluator determines if an eye image from the iris image acquisition module has a quality which is too low for a non-adaptive iris matching module and is sufficiently high enough for the adaptive iris matching module; andwherein if the quality is high enough for the adaptive iris matching module, the adaptive iris matching module executes a search and generates a list of possible matches without segmenting a circumferential of an iris of the eye image.

2. The system of claim 1, wherein the non-adaptive iris matching module processes an iris image having sufficient quality so that a best match for the iris image can be generated from a database of iris images.

3. The system of claim 2, wherein the adaptive iris matching module processes an iris image having low but sufficient quality so that a group of most probable matches for the iris image instead of one best match can be generated from a database of iris images.

4. The system of claim 3, wherein the image quality evaluator determines whether an iris image has sufficient quality for successful processing by the non-adaptive iris matching module or not sufficient quality for successful processing by the non-adaptive iris matching module but has sufficient quality for processing by the adaptive iris matching module, or not sufficient quality for processing by the non-adaptive iris matching module or by the adaptive iris matching module.

5. The system of claim 4, wherein: if an iris image has not sufficient quality for processing by the non-adaptive iris matching module or by the adaptive iris matching module, then the image is rehabilitated, if possible, for processing by the non-adaptive iris matching module or by the adaptive iris matching module; andif the iris image cannot be rehabilitated, then the iris image is rejected.

6. The system of claim 1, wherein the image quality evaluator determines a quality of an iris image relative to blur, obscuration, standoff angle, and/or iris visibility of the iris image.

7. The system of claim 1, wherein the adaptive iris matching module is for segmenting a visible portion of an iris image, partial encoding, database indexing, and/or adaptive matching relative to the visible portion of the iris image.

8. The system of claim 7, wherein the adaptive iris matching module is further for normalizing data points of an iris portion along a visible radial at an angle segment of an iris image.

9. The system of claim 7, wherein the adaptive iris matching module is further for encoding a partial visible iris map signature of an iris image.

10. The system of claim 7, wherein the database indexing comprises: providing a database having templates; andgrouping the database into a plurality of sub-databases; andwherein:each template in the database has a first number of bits based on angular indexing;each sub-database of the plurality of sub-databases represents virtually all templates in the database; and/oreach template in a sub-database has a second number of bits based on a hierarchical structure.

11. The system of claim 10, wherein the database indexing further comprises: providing a subset of template barcodes for possible matching; and/orselecting a second number of bits of a barcode that corresponds to the second number of bits of each template.

12. The system of claim 11, wherein the adaptive iris matching module is further for normalizing a number of data points from an obscured iris portion of an iris image to match one of the templates.

13. The system of claim 12, wherein the adaptive iris matching module is further for encoding a partial obscured iris map signature.

14. The system of claim 13, wherein the adaptive iris matching module is further for comparing iris barcodes to a previously generated reference iris signature.

15. A method for standoff iris recognition, comprising: segmenting a visible region of an iris, with a model providing a first order estimate of a border segment;extracting associated partial codes of a barcode of the iris to index a smaller population of a whole population; andrecovering noisy information in an obscured area by adapting a fitting model that provides best matching bits within the visible region of the iris;wherein:iris dyadic variations are captured by formulating a dynamic encoder that takes into account varying edge points; andan encoding scheme is modified to make use of a dynamic wavelet function wherein a wavelength scalar of the wavelet function varies as a function of an angular variable.

16. The method of claim 15, wherein: the dynamic wavelet function is a smooth continuous function of the angular variable; andan output of the dynamic wavelet function is enabled at each angle using virtually all possible values for a best match.

17. The method of claim 16, wherein concatenated matched bits can determine an overall match score between a probe and a query barcode.

18. A method for adaptive iris matching, comprising: acquiring an image of an iris;evaluating quality of the image;processing a visible region of the iris in the image on a radial direction without segmenting a circumferential of the iris in the image, if the quality is too low for non-adaptive iris matching but high enough for adaptive iris matching;estimating a boundary for the visible region of the iris in the image;constructing an iris map using a rubber sheet model;masking a non-visible region of the iris;encoding the iris map for at least partial comparisons between templates;extracting associated partial codes of an iris code to index a portion of a database; andretrieving a list of two or more possible iris matches from the database with the iris code as a query.