Method and system for recognizing faces

1. A method for recognizing faces, comprising:

retrieving a pair of face images;

segmenting each of the retrieved face images into a plurality of image patches, wherein each patch in one image and a corresponding one in the other image form a pair of patches;

determining a first similarity of each pair of patches;

determining, from all pair of patches, a second similarity of the pair of face images; and

fusing the first similarity determined for the each pair of patches and the second similarity determined for the pair of face images;

wherein determining, from all pair of patches, a second similarity comprises:

obtaining each of pair of patches from the plurality of image patches;

obtaining a plurality of adjacent patches surrounding the obtained patches in a second KNN;

retrieving, from the second KNN, sub networks for the adjacent patches; and

determining the second similarity in the retrieved subnetworks.

2. The method according to claim 1 , wherein determining a first similarity of each pair of patches comprises:

obtaining each of the pair of patches and K adjacent patches surrounding the obtained patches, where K is integer and more than 1;

forming a first KNN from the obtained patches; and

determining the first similarity of each pair of patches in the formed first KNN.

3. The method according to claim 1 , wherein the first similarity is determined by performing random walks in the formed first KNN.

4. The method according to claim 1 , further comprising forming the second KNN, which comprises:

retrieving N training face images, where N is integer and more than 1;

segmenting each of the training face images into M patches of image;

forming M KNNs from the segmented patches for M image patches of N training face images;

linking each patch to its adjacent patches in each of formed KNN, so as to form the second KNN.

5. The method according to claim 1 , wherein the second similarity is determined by performing random walks in the network.

6. A system for recognizing face, comprising:

a segmenting unit configured to retrieve a pair of face images and segment each of the face images into a plurality of patches of image, each patch in one image and a corresponding one in the other image form a pair of patches;

a first similarity determining unit configured to determine a first similarity of each pair of patches;

a second similarity determining unit configured to determine, from all pair of patches, a second similarity of the pair of face images; and

a fusing unit configured to fuse the first similarity determined for the each pair of patches and the second similarity determined for the pair of face images;

wherein the second similarity determining unit is configured to:

obtain each of pair of patches from the plurality of image patches;

obtain a plurality of adjacent patches surrounding the obtained patches in a second KNN;

retrieve, from the second KNN, sub networks for the adjacent patches; and

determine the second similarity in the retrieved sub-networks.

7. The system according to claim 6 , wherein the first similarity determining unit is configured to,

obtain each of the pair of patches and K adjacent patches surrounding the obtained patches, where K is integer and more than 1;

form a first KNN from the obtained patches; and

determine the first similarity of each pair of patches in the formed first KNN.

8. The system according to claim 6 , wherein the first similarity is determined by performing random walks in the formed first KNN.

9. The system according to claim 6 , wherein the second similarity determining unit operates to form the second KNN by:

retrieving N training face images, where N is integer and more than 1;

segmenting each of the training face images into M patches of image;

forming M KNNs from the segmented patches for M image patches of N training face images;

linking each patch to its adjacent patches in each of formed KNN, so as to formed the second KNN.

10. The system according to claim 6 , wherein the second similarity is determined by performing random walks in the network.

11. A non-transitory computer readable storage medium, comprising:

instructions for retrieving a pair of face images;

instructions for segmenting each of the retrieved face images into a plurality of image patches, wherein each patch in one image and a corresponding one in the other image form a pair of patches;

instructions for determining a first similarity of each pair of patches;

instructions for determining, from all pair of patches, a second similarity of the pair of face images; and

instructions for fusing the first similarity determined for the each pair of patches and the second similarity determined for the pair of face images;

wherein the instructions for determining, from all pair of patches, a second similarity of the pair of face images comprise:

instructions for obtaining each of pair of patches from the plurality of image patches;

instructions for obtaining a plurality of adjacent patches surrounding the obtained patches in a second KNN;

instructions for retrieving, from the second KNN, sub networks for the adjacent patches; and

instructions for determining the second similarity in the retrieved subnetworks.

12. The non-transitory computer readable storage medium according to claim 11 , wherein the instructions are included in a computer program product embodied on the non-transitory computer readable storage medium.

13. The non-transitory computer readable storage medium according to claim 11 , wherein the non-transitory computer readable storage medium is disk storage, CD-ROM or optical memory.