Back to results
US 8,577,095 B2Granted
System and method for non-cooperative iris recognition
AssigneeIndiana University Research and Technology Corporation
InventorsEliza Yingzi Du, Craig Belcher
Life of this patent
Estimating the term…
Abstract
A method processes segmented iris images obtained by a non-cooperative image acquisition system to generate descriptors for features in the segmented iris image that are tolerant of segmentation error. The method includes receiving a segmented iris image, and selecting feature points in the segmented iris image to describe an iris locally.
Claims14
A method for processing segmented iris images comprising:
- receiving a segmented iris image;
- dividing an iris region in the segmented iris image into a plurality of bins.
- dividing the iris region in the segmented iris image into a second plurality of bins angularly offset from the first plurality of bins to include iris features at edges of bins in the first plurality of bins;
- selecting feature points in the segmented iris image to describe an iris locally;
- comparing selected feature points in a first location in the segmented iris image to feature points at a corresponding location in another iris image;
- and
- detecting a correspondence between the segmented iris image and the other iris image in response to the Euclidian distance between the selected feature points in the first location in the segmented iris image and the feature points at the corresponding location in the other iris image being less than a predetermined threshold.
4 more elements
A method for processing iris images comprising:
- receiving a first image corresponding to an iris of an eye having a first dilation and a second image corresponding to the iris of the eye having a second dilation;
- generating a first plurality of bins in a rectangular arrangement, each bin in the first plurality of bins corresponding to a predetermined region in the first image;
- generating a second plurality of bins in the rectangular arrangement, each bin in the second plurality of bins corresponding a predetermined region in the second image, at least one of the predetermined regions in the second image having a different size than a corresponding predetermined region in the first image and each bin in the second plurality of bins corresponding to one bin in the first plurality of bins;
- identifying at least one stable feature in one bin of the first plurality of bins and a corresponding bin in the second plurality of bins;
- generating a first feature descriptor with a Gabor filter applied to the one bin in the first plurality of bins and a plurality of bins proximate to the one bin in the first plurality of bins;
- generating a second feature descriptor with the Gabor filter applied to the corresponding bin in the second plurality of bins and another plurality of bins proximate to the second bin in the second plurality of bins;
- identifying a Euclidean distance between the first feature descriptor and the second feature descriptor;
- and
- identifying that the first image and the second image correspond to a single iris in response to the identified Euclidean distance being less than a predetermined threshold.
6 more elements
Dependencies are inferred from the claim text.