Texture features for biometric authentication
One aspect comprises obtaining one or more image regions from a first image of an eye. Each of the image regions may include a view of a respective portion of the white of the eye. The aspect may further comprise applying several distinct filters to each of the image regions to generate a plurality of respective descriptors for the region. The several distinct filters may include convolutional filters that are each configured to describe one or more aspects of an eye vasculature and in combination describe a visible eye vasculature in a feature space. A match score may be determined based on the generated descriptors and based on one or more descriptors associated with a second image of eye vasculature.
1. A computer implemented method comprising:
obtaining one or more first image regions of a first image of an eye, wherein each of the first image regions includes a view of a respective portion of vasculature of the eye;
training a classifier using a training algorithm and descriptors derived from filters applied to eye images;
determining a confidence threshold of the trained classifier based on a false reject or a false accept point of a receiver operating characteristic curve generated from a sensitivity analysis of the classifier; and
until a score exceeds the confidence threshold, progressively increasing a filter resolution and calculating the score based on the filter resolution as follows:
applying a plurality of filters having the filter resolution to the first image regions to generate filter output;
deriving a plurality of first descriptors from the filter output;
defining a plurality of second descriptors derived from second image regions of a second image of an eye, the second image regions comprising regions that are co-registered with the first image regions for corresponding descriptors in the plurality of first descriptors; and
providing the first descriptors and at least a subset of the second descriptors as input to the classifier and obtaining the score as output of the classifier.
2. The method of claim 1 wherein applying the plurality of filters comprises convolving the filters with the first image regions to generate the filter output.
3. The method of claim 2 wherein convolving the filters with the first image regions comprises multiplying frequency domain representations of the first image regions with the filters' kernels and then transforming the results to the spatial domain to generate the filter output.
4. The method of claim 1 wherein the filters are Gabor filters.
5. The method of claim 4 wherein each of the filters has a different orientation.
6. The method of claim 1 wherein a particular first descriptor is an average or median magnitude of image data elements of a portion of the filter output.
7. The method of claim 1 wherein a particular first descriptor is an average or median phase of image data elements of a portion of the filter output.
8. The method of claim 1 wherein the score is calculated based on a distance between the first descriptors and the at least a subset of the second descriptors.
9. The method of claim 8 wherein the distance is a Euclidian distance, a correlation coefficient, a modified Hausdorff distance, a Mahalanobis distance, a Bregman divergence, a cosine similarity, a Kullback-Leibler distance, or a Jensen-Shannon divergence.
10. A system comprising:
data processing apparatus programmed to perform operations comprising:
obtaining one or more first image regions of a first image of an eye, wherein each of the first image regions includes a view of a respective portion of vasculature of the eye;
training a classifier using a training algorithm and descriptors derived from filters applied to eye images;
determining a confidence threshold of a trained classifier based on a false reject or a false accept point of a receiver operating characteristic curve generated from a sensitivity analysis of the classifier;
until a score exceeds the confidence threshold, progressively increasing a filter resolution and calculating the score based on the filter resolution as follows:
applying a plurality of filters having the filter resolution to the first image regions to generate filter output;
deriving a plurality of first descriptors from the filter output;
defining a plurality of second descriptors derived from second image regions of a second image of an eye, the second image regions comprising regions that are co-registered with the first image regions for corresponding descriptors in the plurality of first descriptors; and
providing the first descriptors and at least a subset of the second descriptors as input to the classifier and obtaining the score as output of the classifier.
11. The system of claim 10 wherein applying the plurality of filters comprises convolving the filters with the first image regions to generate the filter output.
12. The system of claim 11 wherein convolving the filters with the first image regions comprises multiplying frequency domain representations of the first image regions with the filters' kernels and then transforming the results to the spatial domain to generate the filter output.
13. The system of claim 10 wherein the filters are Gabor filters.
14. The system of claim 13 wherein each of the filters has a different orientation.
15. The system of claim 10 wherein a particular first descriptor is an average or median magnitude of image data elements of a portion of the filter output.
16. The system of claim 10 wherein a particular first descriptor is an average or median phase of image data elements of a portion of the filter output.
17. The system of claim 10 wherein the score is calculated based on a distance between the first descriptors and the at least a subset of the second descriptors.
18. The system of claim 17 wherein the distance is a Euclidian distance, a correlation coefficient, a modified Hausdorff distance, a Mahalanobis distance, a Bregman divergence, a cosine similarity, a Kullback-Leibler distance, or a Jensen-Shannon divergence.
19. A program product stored on a non-transitory computer-readable medium, the program product comprising instructions that, when executed by data processing apparatus, cause the data processing apparatus to perform operations comprising:
obtaining one or more first image regions of a first image of an eye, wherein each of the first image regions includes a view of a respective portion of vasculature of the eye;
training a classifier using a training algorithm and descriptors derived from filters applied to eye images;
determining a confidence threshold of a trained classifier based on a false reject or a false accept point of a receiver operating characteristic curve generated from a sensitivity analysis of the classifier;
until a score exceeds the confidence threshold, progressively increasing a filter resolution and calculating the score based on the filter resolution as follows:
applying a plurality of filters having the filter resolution to the first image regions to generate filter output;
deriving a plurality of first descriptors from the filter output;
defining a plurality of second descriptors derived from second image regions of a second image of an eye, the second image regions comprising regions that are co-registered with the first image regions for corresponding descriptors in the plurality of first descriptors; and
providing the first descriptors and at least a subset of the second descriptors as input to the classifier and obtaining the score as output of the classifier.
20. The program product of claim 19 wherein applying the plurality of filters comprises convolving the filters with the first image regions to generate the filter output.
21. The program product of claim 20 wherein convolving the filters with the first image regions comprises multiplying frequency domain representations of the first image regions with the filters' kernels and then transforming the results to the spatial domain to generate the filter output.
22. The program product of claim 19 wherein the filters are Gabor filters.
23. The program product of claim 22 wherein each of the filters has a different orientation.
24. The program product of claim 19 wherein a particular first descriptor is an average or median magnitude of image data elements of a portion of the filter output.
25. The program product of claim 19 wherein a particular first descriptor is an average or median phase of image data elements of a portion of the filter output.
26. The program product of claim 19 wherein the score is calculated based on a distance between the first descriptors and the at least a subset of the second descriptors.
27. The program product of claim 26 wherein the distance is a Euclidian distance, a correlation coefficient, a modified Hausdorff distance, a Mahalanobis distance, a Bregman divergence, a cosine similarity, a Kullback-Leibler distance, or a Jensen-Shannon divergence.