IRIS BIOMETRIC MATCHING SYSTEM

• US 20140161325A1
• Filed: 12/09/2013
• Published: 06/12/2014
• Est. Priority Date: 12/10/2012
• Status: Active Grant

• Current Assignee: Princeton Identity Inc.
• Original Assignee: SRI International, Inc.
• Inventors: Bergen, James R.
• Granted Patent: US 9,195,890 B2
• Time in Patent Office: Days
• US Class Current: 382/117
• CPC Class Codes: G06T 7/12 Edge-based segmentation
• CPC Class Codes: G06V 40/10 Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
• CPC Class Codes: G06V 40/193 Preprocessing; Feature extraction
• CPC Class Codes: G06V 40/197 Matching; Classification

Abstract

A method and apparatus for biometric iris matching comprising pre-processing an input image capturing one or more eyes to produce one or more rectified iris images, coding the one or more rectified iris images into one or more multiresolution iris codes and matching the one or more multiresolution iris code with a set of stored multiresolution iris codes to determine whether a match exists.

Claims

1. A method for biometric iris matching comprising:

pre-processing an input image capturing one or more eyes to produce one or more rectified iris images;

coding the one or more rectified iris images into one or more multiresolution iris codes; and

matching the one or more multiresolution iris code with a set of stored multiresolution iris codes to determine whether a match exists.

2. The method of claim 1 wherein the pre-processing further comprises:

segmenting a pupil area in each of the one or more eyes in the input image;

segmenting an iris area from a sclera area in each of the one or more eyes in the input image; and

inserting an artificial pupil within the segmented pupil area.

3. The method of claim 2 wherein segmenting the pupil and segmenting an iris/sclera boundary comprises:

generating one or more edge maps corresponding to each eye in the input image;

constructing one or more candidate pupil contours based on the one or more edge map, wherein each edge map contains at least one candidate pupil contour; and

determining a best matching candidate pupil contour for each eye in the input image.

4. The method of claim 3 wherein constructing one or more candidate pupil contours comprises:

constructing a first candidate pupil contour from a best fitting circle for each of the one or more edge maps;

constructing a second candidate pupil contour from a best inscribed circle for each of the one or more edge maps; and

determining the best matching candidate pupil contour between the first candidate pupil contour and the second candidate pupil contour.

5. The method of claim 4 further comprising:

determining the best match by assessing a level of support for the best fitting circle and selecting the best fitting circle as the best match if this level of support is above a threshold value, and selecting the best inscribed circle as the best match if the level of support for the best fitting circle is below a threshold value.

6. The method of claim 1 wherein pre-processing further comprises:

estimating an angle of tilt of the eye in the input image with respect to a camera that captured the image;

correcting for perspective distortion based on the estimated angle of tilt; and

correcting corneal distortion based on the estimated angle of tilt by aligning an iris contour and a pupil contour of the eye into concentric circles.

7. The method of claim 6 wherein correcting for perspective distortion comprises one of:

stretching the input image in a direction of the estimated angle of tilt by applying an affine or projective transformation to the input image.

8. The method of claim 6 wherein correcting corneal distortion comprises:

defining a plurality of spokes connecting a plurality of points on the iris contour to a plurality of points on the pupil contour;

mapping the plurality of spokes to a position connecting the plurality of points on the iris contour to a synthetic circular pupil contour; and

applying the mapping to the input image.

9. The method of claim 1, wherein coding the one or more rectified irises further comprises:

transforming the rectified iris images into transformed iris images, wherein the transformed iris images are invariant coordinate system image representations of the rectified iris images;

masking off specular reflections and unstable regions in the transformed iris images to produce masked converted iris images; and

subsampling the masked transformed iris images using a bandpass filter into a plurality of filtered images, wherein the plurality of filtered images range from high frequency bands to low frequency bands and the plurality of bands comprise an iris code.

10. The method of claim 9 further comprising:

subsampling the masked transformed iris images to produce five frequency band images, wherein the bands comprise image information with approximate angular spatial frequencies of 33, 25, 16, 12 and 8 cycles per two times pi radians.

11. The method of claim 1, wherein matching the one or more iris codes further comprising:

aligning a coded iris code and a stored iris code with each other to ease matching;

determining a Hamming Distance between the coded iris code and the stored iris code; and

determining that a match exists when the Hamming Distance is less than a predetermined threshold value.

12. The method of claim 11 wherein aligning further comprises:

computing a local displacement function between the coded iris code and the stored iris code using barrel shift alignment.

13. The method of claim 11 wherein aligning further comprises:

computing a flow field estimation between the coded iris code and the stored iris code, where vectors in the flow field each specify displacement in a normalized image coordinate system.

14. The method of claim 13, further comprising:

computing the flow field for each band within the coded iris code and the stored iris code.

15. The method of claim 13, wherein computing the flow field further comprises:

generating a first plurality of images from a multiresolution code corresponding to the input image and generating a second plurality of images from a multiresolution code corresponding to a stored image using a bandpass filter, the first and second plurality of images ranging from low frequency to high frequency bands;

selecting an image from the first plurality of images in a lowest frequency band that has not been processed;

determining whether a flow field has been estimated in a lower frequency band between the first and second plurality of images;

warping the selected image if a lower frequency band flow field is estimated using a flow field estimate of the lower frequency band;

estimating a flow field on residual difference between the warped image or selected image and a second image at a same frequency level from the second plurality of images; and

repeating the above steps until all frequency bands in the first plurality of images have been processed and a final flow field is estimated.

16. The method of claim 15 further comprising:

using the frequency band images in the coded iris code as the first plurality of images, and the frequency band images in the stored iris code as the second plurality of images.

17. The method of claim 11 further comprising wherein the predetermined threshold value corresponds to details of a structure of the iris codes and on statistical requirements of a matching scenario.

18. The method of claim 11 further comprising:

determining a Hamming Distance for each band between the coded iris code and the stored iris code; and

determining a matching score for each band and determining whether a match exists based on a combination of matching scores for each band.

19. The method of claim 2, wherein coding the one or more rectified irises further comprises:

transforming the rectified iris images into a transformed iris images, wherein the transformed iris images are invariant coordinate system image representations of the rectified iris images;

masking off specular reflections and unstable regions in the transformed iris images to produce masked converted iris images; and

subsampling the masked transformed iris images using a bandpass filter into a plurality of filtered images, wherein the plurality of filtered images range from high frequency bands to low frequency bands and the plurality of band images comprise an iris code.

20. An apparatus for biometric iris matching comprising:

a pre-processor that pre-processes an input image capturing one or more eyes to produce one or more rectified iris images;

a coding processor that codes the one or more rectified iris images into one or more multiresolution iris codes; and

a matching processor that matches the one or more multiresolution iris code with a set of stored multiresolution iris codes to determine whether a match exists.

21. A method for distinguishing a pupil comprising:

segmenting a pupil area in one or more eyes detected in an input image; and

segmenting an iris area from a sclera area in each of the one or more eyes in the input image.

22. The method of claim 21 wherein segmenting the pupil and iris/sclera boundary comprises:

generating one or more edge maps corresponding to each eye in the input image;

constructing one or more candidate pupil contours based on the one or more edge map, wherein each edge map contains at least one candidate pupil contour; and

determining a best matching candidate pupil contour for each eye in the input image.

23. The method of claim 22 wherein constructing one or more candidate pupil contours comprises:

constructing a first candidate pupil contour from a best fitting circle for each of the one or more edge maps;

constructing a second candidate pupil contour from a best inscribed circle for each of the one or more edge maps; and

determining the best matching candidate pupil contour between the first candidate pupil contour and the second candidate pupil contour.

24. A method for estimating tilt of an eye in an input image comprising:

receiving an orientation of a camera capturing one or more frames of an eye;

and estimating a direction of a tilt of the eye with respect to the camera orientation.

25. The method of claim 24 wherein correcting for perspective distortion comprises one of:

stretching the input image in the estimated direction of tilt and applying an affine or projective transformation to the input image.

26. The method of claim 24 wherein correcting corneal distortion comprises:

defining a plurality of spokes connecting a plurality of points on an iris contour to a plurality of points on a pupil contour of the eye;

mapping the plurality of spokes to a position connecting the plurality of points on the iris contour to a synthetic circular pupil contour; and

applying the mapping to the input image.