Face matching for mobile devices
A method includes receiving an image of a face to match with images of known faces, extracting blocks multiple blocks from the received image, calculating local binary pattern histograms for each block, generating matching scores for each block against block of the images of known faces, determining a top number, N, of matching scores less than the number of blocks, and matching the received image to an image of a known face as a function of the top number of matching scores.
1. A method comprising:
receiving an image of a face to match with images of known faces;
extracting multiple blocks from the received image;
calculating local binary pattern histograms for each block;
generating matching scores for each block against block of the images of known faces;
determining a top number, N, of sorted matching scores less than the number of blocks; and
matching the received image to an image of a known face as a function of the top number of matching scores.
2. The method of claim 1 and further comprising granting access based on the match.
3. The method of claim 1 and further comprising aligning the received image based on a landmark localization of the face in the image prior to dividing the received image into blocks.
4. The method of claim 3 and further comprising determining an entropy measure to identify misalignment of the image of the face.
5. The method of claim 4 and further comprising introducing multiple images of a known face with perturbed eye locations to improve accuracy of the matching.
6. The method of claim 5 and further comprising:
searching among the multiple images for a known face by selecting a best horizontal difference for one eye, selecting a best vertical difference from the best horizontal difference; and for the best eye position, selecting the image with the best other eye position.
7. The method of claim 4 wherein the top number of matching scores is determined as a function of the entropy measure.
8. The method of claim 1 wherein the blocks include dividing the receiving image into multiple non-overlapping blocks.
9. The method of claim 1 wherein the blocks include dividing the receiving image into multiple overlapping blocks.
10. The method of claim 1 wherein the blocks include dividing the receiving image into different size blocks.
11. The method of claim 1 wherein matching is a function of the sum of the N scores.
12. The method of claim 1 wherein N is determined as a function resolution of the received image.
13. The method of claim 1 wherein N is determined as a function of landmark localization error.
14. A non-transitory computer readable storage device having instructions for causing a computer to perform a method, the method comprising:
receiving an image of a face to match with images of known faces;
extracting multiple blocks from the received image;
calculating local binary pattern histograms for each block;
generating matching scores for each block against block of the images of known faces;
determining a top number, N, of sorted matching scores less than the number of blocks; and
matching the received image to an image of a known face as a function of the top number of matching scores.
15. A system comprising:
a processor; and
a non-transitory storage device coupled to the processor to store software modules for execution on the processor, the software modules comprising:
a module to receive an image of a face to match with images of known faces;
a module to extract multiple blocks from the received image;
a module to calculate local binary pattern histograms for each block;
a module to generate matching scores for each block against block of the images of known faces;
a module to determine a top number, N, of sorted matching scores less than the number of blocks; and
a module to match the received image to an image of a known face as a function of the top number of matching scores.
16. The system of claim 13 and further comprising a database to store and provide access to the images of known faces.