Method for localizing irises in images using gradients and textures

• US 20070160267A1
• Filed: 01/11/2006
• Published: 07/12/2007
• Est. Priority Date: 01/11/2006
• Status: Active Grant

• Current Assignee: Mitsubishi Electric Research Laboratories, Inc. (Mitsubishi Electric Corporation)
• Original Assignee: Mitsubishi Electric Research Laboratories, Inc. (Mitsubishi Electric Corporation)
• Inventors: Jones, Michael, Guo, Guodong
• Granted Patent: US 7,583,823 B2
• Time in Patent Office: Days
• US Class Current: 382/117
• CPC Class Codes: G06V 40/193 Preprocessing; Feature extraction

Abstract

A method for extracting an iris from an image is presented. Edges of an iris are detected in an image. Texture from the image is acquired. The edges and texture are combined to generate an inner boundary and an outer boundary of the iris. Also, to improve the accuracy of the iris boundary detection, a method to select between ellipse and circle models is provided. Finally, a dome model is used to determine mask images and remove eyelid occlusions in unwrapped images.

Claims

1. A method for localizing pixels representing an iris in an image acquired of an eye, comprising:

determining gradient information from an image acquired of an eye;

determining texture information from the image; and

combining the gradient information and the texture information to generate boundaries of an iris image including pixels representing an iris in the image of the eye.

2. The method of claim 1, in which the gradient information is determined using an integro-differential operator.

3. The method of claim 1, in which the determining of the texture information further comprises:

defining an inner zone of pixels and an outer zone of pixels in the image;

constructing histograms to represent the pixels in the inner zone and the outer zone; and

measuring a difference between the histograms to determine the texture information.

4. The method of claim 3, in which the constructing further comprises:

determining values of the pixels using a local binary pattern operator.

5. The method of claim 3, in which the difference is a Kullback-Leibler divergence.

6. The method of claim 1, in which the boundaries include an inner boundary and an outer boundary.

7. The method of claim 6, further comprising:

unwrapping the iris image according to the inner boundary and the outer boundary to produce an unwrapped iris image;

detecting occlusions in the unwrapped iris image; and

generating a mask for each occlusion.

8. The method of claim 7, further comprising:

defining a first region of the unwrapped image where an occlusion cannot occur;

defining a second region of the unwrapped image where an occlusion may occur;

constructing histograms to represent the pixels in the first and second regions;

determining a distance between the histograms; and

generating a mask in the second region only if the distance is greater than a predetermined threshold.

9. The method of claim 8, in which the distance is a chi-square distance.

10. The method of claim 7, in which the mask is generated according to the edge information and the texture information.

11. A method for localizing pixels representing an iris in an image acquired of an eye, comprising:

measuring a gradient magnitude of pixel intensities in an image of an eye;

determining a texture difference in the image;

combining the gradient magnitude and the texture difference;

defining boundaries of an iris image, including pixels representing an iris in the image, according to the combination of the gradient and the texture difference.

12. The method of claim 11, in which a particular boundary is selected using a set of circles, the selecting further comprising:

repeating the measuring and the determining for each circle in a set of circles;

combining the gradient information and the texture difference for each circle; and

assigning a circle in the set of circles as the boundary, the circle having a maximum of the combination of the gradient magnitude and the texture difference.

13. The method of claim 12, in which the centers of the circles in the set of circles are approximate to a center of a pupil in the image of an eye.

14. The method of claim 11, in which the boundaries are elliptical.

15. The method of claim 12, further comprising,

repeating the measuring, determining, and combining in a region approximate to the boundary to generate contours;

fitting an ellipse to a longest contour; and

reassigning the ellipse as the boundary if an eccentricity of the ellipse is greater than a threshold, else keeping the circle as the boundary.

16. The method of claim 15, in which the threshold is 0.19.