Systems and methods for spoof detection based on local interest point locations

1. A device, comprising:

a biometric sensor; and

a processing system configured to:

receive, from the biometric sensor, an input image of a biometric;

obtain keypoint locations of keypoints in the input image, wherein keypoints in the input image comprise local regions of interest in the input image;

generate a scaled image by scaling the input image to a smaller size;

determine that two or more keypoints in the input image comprise a cluster of keypoints in response to determining that keypoint locations of the two or more keypoints in the input image correspond to a same pixel in the scaled image;

compute one or more anti-spoof metrics of the input image based on the keypoint locations, wherein the one or more anti-spoof metrics comprise a metric corresponding to a number of clusters of keypoints in the input image; and

determine whether the input image is a replica of the biometric based on the one or more anti-spoof metrics.

2. The device of claim 1 , wherein the keypoint locations are obtained from a matcher configured to determine whether the input image is a match to one or more enrollment images separate from the processing system determining whether the input image is a replica of the biometric.

3. The device of claim 2 , wherein the matcher is configured to determine an overlap region of the input image and a first enrollment image, and the one or more anti-spoof metrics further comprise a metric corresponding to a number of keypoints in the overlap region of the input image and the first enrollment image.

4. The device of claim 2 , wherein the matcher is configured to determine an overlap region of the input image and a first enrollment image and a non-overlap region of the input image and the first enrollment image, and the one or more anti-spoof metrics further comprise a metric corresponding to a number of keypoints in the non-overlap region of the input image and the first enrollment image.

5. The device of claim 1 , wherein the one or more anti-spoof metrics comprise a metric corresponding to an average distance score of the keypoints in the input image, wherein a distance score for a given keypoint is computed by calculating an average distance between the keypoint and a number of its closest neighboring keypoints.

6. The device of claim 1 , wherein the keypoint locations are obtained based on the processing system executing a BRISK (Binary Robust Invariant Scalable Keypoints) algorithm on the input image.

7. The device of claim 1 , wherein the processing system is further configured to:

receive an anti-spoof template corresponding to the biometric; and

compute the one or more anti-spoof metrics based on comparing anti-spoof metrics included in the anti-spoof template with the one or more anti-spoof metrics of the input image.

8. The device of claim 1 , wherein the one or more anti-spoof metrics comprise a metric corresponding to a number of keypoints in the input image.

9. The device of claim 1 , wherein the one or more anti-spoof metrics comprise a metric corresponding to an average number of keypoints in the input image based on a size of the input image.

10. A method for performing spoof detection, comprising:

receiving, by processor from a biometric sensor, an input image of a biometric;

obtaining, by the processor, keypoint locations of keypoints in the input image, wherein keypoints in the input image comprise local regions of interest in the input image;

generating a scaled image by scaling the input image to a smaller size;

determining that two or more keypoints in the input image comprise a cluster of keypoints in response to determining that keypoint locations of the two or more keypoints in the input image correspond to a same pixel in the scaled image;

computing, by the processor, one or more anti-spoof metrics of the input image based on the keypoint locations, wherein the one or more anti-spoof metrics comprise a metric corresponding to a number of clusters of keypoints in the input image; and

determining, by the processor, whether the input image is a replica of the biometric based on the one or more anti-spoof metrics.

11. The method of claim 10 , wherein the keypoint locations are obtained from a matcher configured to determine whether the input image is a match to one or more enrollment images separate from the processor determining whether the input image is a replica of the biometric, wherein the matcher is configured to determine an overlap region of the input image and a first enrollment image, and the one or more anti-spoof metrics comprise a metric corresponding to a number of keypoints in the overlap region of the input image and the first enrollment image.

12. The method of claim 10 , wherein the keypoint locations are obtained from a matcher configured to determine whether the input image is a match to one or more enrollment images separate from the processor determining whether the input image is a replica of the biometric, wherein the matcher is configured to determine an overlap region of the input image and a first enrollment image and a non-overlap region of the input image and the first enrollment image, and the one or more anti-spoof metrics comprise a metric corresponding to a number of keypoints in the non-overlap region of the input image and the first enrollment image.

13. The method of claim 10 , wherein the one or more anti-spoof metrics comprise one or more of:

metric corresponding to a number of keypoints in the input image;

a metric corresponding to an average number of keypoints in the input image based on a size of the input image; and

a metric corresponding to an average distance score of the keypoints in the input image, wherein a distance score for a given keypoint is computed by calculating an average distance between the keypoint and a number of its closest neighboring keypoints.

14. The method of claim 10 , wherein the keypoint locations are obtained based on the processor executing a BRISK (Binary Robust Invariant Scalable Keypoints) algorithm on the input image.

15. The method of claim 10 , further comprising:

receiving an anti-spoof template corresponding to the biometric; and

computing the one or more anti-spoof metrics based on comparing anti-spoof metrics included in the anti-spoof template with the one or more anti-spoof metrics of the input image.

16. A non-transitory computer-readable storage medium storing instructions that, when executed by a processor, causes a computing device to perform spoof detection, by performing steps comprising:

receiving, from a biometric sensor, an input image of a biometric;

obtaining keypoint locations of keypoints in the input image, wherein keypoints in the input image comprise local regions of interest in the input image;

generating a scaled image by scaling the input image to a smaller size;

determining that two or more keypoints in the input image comprise a cluster of keypoints in response to determining that keypoint locations of the two or more keypoints in the input image correspond to a same pixel in the scaled image;

computing one or more anti-spoof metrics of the input image based on the keypoint locations, wherein the one or more anti-spoof metrics comprise a metric corresponding to a number of clusters of keypoints in the input image; and

determining whether the input image is a replica of the biometric based on the one or more anti-spoof metrics.

17. The computer-readable storage medium of claim 16 , wherein the one or more anti-spoof metrics comprise one or more of:

a metric corresponding to a number of keypoints in the input image;

a metric corresponding to an average number of keypoints in the input image based on a size of the input image; and

a metric corresponding to an average distance score of the keypoints in the input image, wherein a distance score for a given keypoint is computed by calculating an average distance between the keypoint and a number of its closest neighboring keypoints.

18. The computer-readable storage medium of claim 16 , wherein the keypoint locations are obtained from a matcher configured to determine whether the input image is a match to one or more enrollment images separate from the processor determining whether the input image is a replica of the biometric, wherein the matcher is configured to determine an overlap region of the input image and a first enrollment image, and the one or more anti-spoof metrics comprise a metric corresponding to a number of keypoints in the overlap region of the input image and the first enrollment image.

19. The computer-readable storage medium of claim 16 , wherein the keypoint locations are obtained from a matcher configured to determine whether the input image is a match to one or more enrollment images separate from the processor determining whether the input image is a replica of the biometric, wherein the matcher is configured to determine an overlap region of the input image and a first enrollment image and a non-overlap region of the input image and the first enrollment image, and the one or more anti-spoof metrics comprise a metric corresponding to a number of keypoints in the non-overlap region of the input image and the first enrollment image.

20. The computer-readable storage medium of claim 16 , wherein the keypoint locations are obtained based on the processor executing a BRISK (Binary Robust Invariant Scalable Keypoints) algorithm on the input image.