Biometric authentication

1. A biometric authentication system, comprising:

a database of digital images of biometrics of known people;

a camera for capturing a digital image of a biometric of an unknown person;

a two-dimensional gradient filter coupled with said database and said camera for deriving a spatial gradient of a digital image; and

a match processor coupled with said two-dimensional gradient filter for determining the extent to which the unknown biometric corresponds with a known biometric, by (i) processing a plurality of pairs of blocks of pixels of two spatial gradients derived by said two-dimensional gradient filter, one spatial gradient derived from a digital image of a biometric of a known person and one spatial gradient derived from the digital image of the biometric of the unknown person, wherein each pair of blocks comprises a block from one of the spatial gradients and a similar block from the other of the spatial gradients, to derive metrics related to the relative placements of the pairs of blocks, and (ii) calculating a probability that the known biometric does not correspond with the unknown biometric, based on the derived metrics.

2. The biometric authentication system of claim 1 wherein said two-dimensional gradient filter derives a spatial gradient by averaging horizontal difference quotients over a first sliding window, and by averaging vertical difference quotients over a second sliding window.

3. The biometric authentication system of claim 2 wherein the horizontal difference quotients have a fixed width, and the length of the first sliding window equals the width of the horizontal difference quotients, and wherein the vertical difference quotients have a fixed width, and the length of the second sliding window equals the width of the vertical difference quotients.

4. The biometric authentication system of claim 1 wherein the biometric is at least one skin texture.

5. The biometric authentication system of claim 1 wherein the biometric is a fingerprint.

6. The biometric authentication system of claim 1 wherein the biometric is a palm print.

7. The biometric authentication system of claim 1 wherein the biometric is a retinal scan.

8. The biometric authentication system of claim 1 wherein the biometric is an iris scan.

9. A computer processor-based method for biometric authentication, comprising:

capturing, by a computer, a digital image, denoted by image F, of a biometric of an unknown person;

retrieving, by the computer, a digital image of a biometric of a known person, denoted by image G, from a database of such images;

deriving, by the computer, a two-dimensional spatial gradient, denoted by grad F, of biometric image F, and a two-dimensional spatial gradient, denoted by grad G, of biometric image G;

for each pixel block B k from a plurality of pixel blocks B 1 , B 2 , . . . , B n , searching, by the computer, for a pixel block C k such that values of grad F over block B k substantially correspond with the values of grad G over block C k ;

determining, by the computer, how many pairs of distinct pixel blocks B k and B m satisfy a criterion that the relative displacement between B k and B m nearly equals the relative displacement between pixel blocks C k and C m ; and

concluding, by the computer, whether or not the unknown biometric matches the known biometric, based on said determining.

10. The method of claim 9 wherein said deriving derives a spatial gradient by averaging horizontal difference quotients over a first sliding window, and by averaging vertical difference quotients over a second sliding window.

11. The method of claim 10 wherein the horizontal difference quotients have a fixed width, and the length of the first sliding window equals the width of the horizontal difference quotients, and wherein the vertical difference quotients have a fixed width, and the length of the second sliding window equals the width of the vertical difference quotients.

12. The method of claim 9 wherein said concluding calculates a probability that the unknown biometric does not match the known biometric, based on the results of said determining.

13. The method of claim 9 wherein the biometric is at least one skin texture.

14. The method of claim 9 wherein the biometric is a fingerprint.

15. The method of claim 9 wherein the biometric is a palm print.

16. The method of claim 9 wherein the biometric is a retinal scan.

17. The method of claim 9 wherein the biometric is an iris scan.

18. A computer-readable storage medium storing program code for causing a computing device:

to capture a digital image, denoted by image F, of a biometric of an unknown person;

to retrieve a digital image of a biometric of a known person, denoted by image G, from a database of such images;

to derive a two-dimensional spatial gradient, denoted by grad F, of palm image F, and a two-dimensional spatial gradient, denoted by grad G, of palm image G;

for each pixel block Bk from a plurality of pixel blocks B 1 , B 2 , . . . , B n , to search for a pixel block C k such that values of grad F over block B k substantially correspond with the values of grad G over block C k ;

to determine how many pairs of distinct pixel blocks B k and B m satisfy a criterion that the relative displacement between B k and B m nearly equals the relative displacement between pixel blocks C k and C m ; and

to conclude whether or not the unknown biometric matches the known biometric, based on the number of pairs of distinct pixel blocks that satisfy the criterion.