System and method for non-cooperative iris image acquisition

• US 8,644,565 B2
• Filed: 07/22/2009
• Issued: 02/04/2014
• Est. Priority Date: 07/23/2008
• Status: Expired due to Fees

• Current Assignee: Indiana University Research & Technology Corporation (Indiana University)
• Original Assignee: Indiana University Research & Technology Corporation (Indiana University)
• Inventors: Du, Eliza Yingzi, Zhou, Zhi, Arslanturk, Emrah
• Primary Examiner(s): Le, Vu
• Assistant Examiner(s): WOLDEMARIAM, AKLILU K
• Application Number:
  US13/055,250
• Publication Number:
  US 20110150334A1
• Time in Patent Office: 1,658 Days
• Field of Search: 382/103, 382/115, 382/254, 382/284
• US Class Current: 382/118
• CPC Class Codes: G06V 40/19 Sensors therefor

Abstract

A method segments iris images from eye image data captured from non-cooperative subjects. The method includes receiving a frame of eye image data, and determining whether a pupil exists in the image by detecting glare areas in the image. Upon finding a pupil, subsequent images are processed with reference to the pupil location and a radius is calculated for the pupil. A k means clustering method and principal component analysis are used to locate pupil boundary points, which are fitted to a conic. Using the pupil boundary, an angular derivative is computed for each frame having a pupil and iris boundary points are fitted to a conic to identify an iris region between the iris boundary and the pupil boundary. Noise data are then removed from the iris region to generate an iris segment. A method for evaluating iris frame quality and iris image segmentation quality is also disclosed.

Claims

1. A method for segmenting an eye image to identify portions of the eye image containing iris pattern data comprising:

receiving a frame of eye image data;

convolving the eye image data with a predetermined matrix of coefficients to produce filtered eye image data, the predetermined matrix of coefficients being configured to produce peak points in the filtered eye image data at positions where glare is not present in the eye image data;

comparing the filtered eye image data to a threshold;

identifying a number of peak points in the filtered eye image data, the identified peak points in the filtered eye image data corresponding to positions of no glare in the eye image data; and

segmenting the eye image data to identify portions of the frame containing iris pattern data only in response to at least one peak point being present in the filtered eye image data.

2. The method of claim 1, the segmenting of the eye image data further comprising:

downsampling the frame of eye image data;

clustering the eye image data in the downsampled frame of eye image data;

calculating principal components and basis vectors for the clustered eye image data; and

detecting whether the frame of eye image data contains a pupil image from the clustered eye image data, the principal components, and basis vectors.

3. The method of claim 2 further comprising:

clustering subsequent frames of eye image data with reference to a calculated mean of clustered data, principal components, and basis vectors computed for a previous frame of eye image data in which a pupil was detected.

4. The method of claim 3 further comprising:

identifying a segment in each frame of clustered eye image data corresponding to a glare area on the iris pattern data;

removing the segment in each frame of clustered eye image data corresponding to the glare area from the clustered image data; and

detecting an edge of a pupil with reference to a first cluster in the clustered image data to identify a pupil location and pupil radius in each frame.

5. The method of claim 4 further comprising:

identifying a region of interest with reference to the pupil location and pupil radius;

computing a gradient for the region of interest;

identifying boundary points for the region of interest from the gradient; and

fitting a conic shape to the boundary points with reference to the pupil location and pupil radius.

6. The method of claim 5, the conic shape being an ellipse.

7. The method of claim 5 further comprising:

computing an intensity value for the region of interest;

comparing the calculated intensity value to intensity values for adjacent regions;

computing an intensity variance for the region of interest;

comparing the intensity variance to a threshold; and

identifying an iris boundary in response to the computed intensity being less than the intensity of the adjacent regions and the computed intensity variance being less than the threshold.

8. The method of claim 7 further comprising:

generating an angular derivative of eye image data in the frame of eye image data;

identifying iris boundary points from the angular derivative; and

fitting a conic shape to the iris boundary points with reference to the pupil location and pupil radius to identify an iris boundary.

9. The method of claim 8, the conic shape being an ellipse.

10. The method of claim 8 further comprising:

computing an iris center for a portion of the eye image data between the pupil boundary and the iris boundary;

computing a center for an area within the pupil boundary;

computing a distance between the iris center and the pupil center;

comparing the computed distance to a threshold; and

removing noise data from the area between the iris boundary and the pupil boundary in response to the computed distance being less than the threshold.

11. The method of claim 10, the noise data removal further comprising:

generating variance values for windows about pixels of eye image data between the iris boundary and the pupil boundary;

comparing the variance values to a threshold; and

removing pixels from an iris region between the iris boundary and the pupil boundary in response to the variance value for the window centered about the pixel being greater than the threshold.

12. The method of claim 11, the noise data removal further comprising:

computing a mean of a gradient for pixels within the windows centered about the pixels in the iris region;

comparing the computed mean to a threshold; and

removing pixels from an iris region in response to the computed gradient mean for the window centered about the pixel being greater than the threshold.

24. The method of claim 1 wherein the predetermined matrix of coefficients is:

13. A method for evaluating quality in a non-cooperative iris image identification system comprising:

identifying regions of glare and motion blur in each frame of a plurality of frames in a video sequence of eye images, the detection of the regions of glare and motion blur further comprising;

applying a Sobel filter to image data in a frame;

identifying bright area pixels corresponding to the Sobel filtered image data exceeding a first threshold;

identifying gray area pixels in the Sobel filtered image data corresponding to the bright area pixels, the gray area pixels being less than a second threshold, the second threshold being greater than the first threshold;

identifying a quality score of the frame with reference to a first number of identified bright area pixels and a second number of identified gray area pixels;

discarding frames from the video sequence for segmentation processing in response to the identified number of gray area pixels being greater than a predetermined percentage of the bright area pixels for the frame;

generating a plurality of iris image segments with reference to frames of the video sequence for which the quality score is greater than the threshold;

generating a segmentation score for the plurality of iris segments, the generation of the segmentation score further comprising;

generating a pupil boundary segmentation score;

generating an iris boundary segmentation score

normalizing the pupil boundary segmentation score and the iris boundary segmentation score to produce the segmentation score; and

generating an overall score for an iris pattern in the frame used to identify a person with reference to a combination of the generated segmentation score and the identified quality score.

14. The method of claim 13, the identification of glare regions and motion blur regions in the video frames further comprising:

detecting interlaced frames in a video sequence.

15. The method of claim 14, the interlaced frame detection further comprising:

calculating correlation coefficients for portions of an image in a frame; and

comparing the correlation coefficients to a non-interlaced frame correlation coefficient.

16. The method of claim 13, the identification of glare regions and motion blur regions in the video frames further comprising:

detecting out of focus frames.

17. The method of claim 16, the out of focus frame detection further comprising:

convolving each frame with a high pass filter;

calculating a sum of squared error for the convolved frame; and

comparing the sum of squared error to a threshold.

18. The method of claim 13, the segmentation score generation further comprising:

evaluating a pupil center and pupil boundary in an iris segmentation.

19. The method of claim 18, the pupil center and pupil boundary evaluation further comprising:

comparing a pupil boundary mask to an iris segmentation mask to measure how well the pupil boundary mask matches the iris segmentation mask.

20. The method of claim 13, the segmentation score generation further comprising:

evaluating an iris boundary in an iris segmentation.

21. The method of claim 20, the iris boundary evaluation further comprising:

calculating a homogeneous measure for an iris boundary in an iris segmentation.

22. The method of claim 21, the homogeneous measure calculation further comprising:

calculating a homogeneous measure for different portions of an iris boundary in an iris segmentation.

23. The method of claim 22, the homogeneous measure calculation further comprising:

generating a composite homogeneous measure from the homogeneous measures calculated for the different iris boundary portions.