METHOD AND APPARATUS FOR EXTRACTING FACE FEATURE

• US 20080130962A1
• Filed: 07/26/2007
• Published: 06/05/2008
• Est. Priority Date: 12/05/2006
• Status: Active Grant

• Current Assignee: Electronics and Telecommunications Research Institute
• Original Assignee: Electronics and Telecommunications Research Institute
• Inventors: CHUNG, Kyo-II, MOON, Ki-Young, LEE, Yongjin
• Granted Patent: US 7,949,158 B2
• Time in Patent Office: Days
• US Class Current: 382/118
• CPC Class Codes: G06V 40/169 Holistic features and representations, i.e. based on the facial image taken as a whole
• CPC Class Codes: G06V 40/172 Classification, e.g. identification

Abstract

A method and apparatus for extracting a face feature to construct a face search system having a high recognition rate are provided. The method includes calculating an average image of each of a plurality of candidates having one or more face images to which different weights are given according to an order in which the face images are acquired and calculating a total average image of all face images of the plurality of candidates, calculating a between-class scatter matrix based on a difference between each candidate'"'"'s average image and the total average image and a within-class scatter matrix based on a difference between each candidate'"'"'s average image and each candidate'"'"'s predetermined face image of the one or more face images, and generating a base vector that maximizes a criterion defined by a ratio between the between-class scatter matrix and the within-class scatter matrix.

Claims

1. A method of extracting a face feature, the method comprising:

calculating an average image of each of a plurality of candidates having one or more face images to which different weights are given according to an order in which the face images are acquired and calculating a total average image of all face images of the plurality of candidates;

calculating a between-class scatter matrix based on a difference between each candidate'"'"'s average image and the total average image and a within-class scatter matrix based on a difference between each candidate'"'"'s average image and each candidate'"'"'s predetermined face image of the one or more face images; and

generating a base vector that maximizes a criterion defined by a ratio between the between-class scatter matrix and the within-class scatter matrix.

2. The method of claim 1, further comprising extracting a face feature by projecting each candidate'"'"'s average image on the base vector.

3. The method of claim 1, wherein the weight has a value of 0 to 1 and has a value close to 1 when the image is more recent.

4. The method of claim 1, wherein the calculating of the between-class scatter matrix and the within-class scatter matrix comprises calculating a between-class scatter without giving a weight according to the number of each candidate'"'"'s face images to prevent the scatter from being biased by the number of the candidate'"'"'s face images and calculating a within-class scatter using normalization using the number of the candidate'"'"'s face images.

5. The method of claim 2, further comprising, when a face image of a new candidate other than the existing candidates is input, extracting a face feature used to register the new candidate by projecting the input face image on the base vector.

6. The method of claim 2, further comprising, when a face image of one candidate among the existing candidates is input:

extracting a face feature used to confirm the identity of the candidate by projecting the input face image on the base vector;

comparing the extracted face feature with face features of the existing candidates; and

outputting a candidate list according to a result of the comparison.

7. The method of claim 5 or 6, further comprising updating previous statistic values which are values calculated in each step based on the previous statistic values and the input face image.

8. An apparatus for extracting a face feature, the apparatus comprising:

an average image calculator calculating an average image of each of a plurality of candidates having one or more face images to which different weights are given according to an order in which the face images are acquired and calculating a total average image of all face images of the plurality of candidates;

a scatter calculator calculating a between-class scatter matrix based on a difference between each candidate'"'"'s average image and the total average image and a within-class scatter matrix based on a difference between each candidate'"'"'s average image and each candidate'"'"'s predetermined face image of the one or more face images; and

a base vector generator generating a base vector that maximizes a criterion defined by a ratio between the between-class scatter matrix and the within-class scatter matrix.

9. The apparatus of claim 8, further comprising a face feature extractor extracting a face feature by projecting each candidate'"'"'s average image on the base vector.

10. The apparatus of claim 8, wherein the weight has a value of 0 to 1 and has a value close to 1 when the image is more recent.

11. The apparatus of claim 8, wherein the scatter calculator calculates a between-class scatter without giving a weight according to the number of each candidate'"'"'s face images to prevent the scatter from being biased by the number of the candidate'"'"'s face images and calculates a within-class scatter using normalization using the number of the candidate'"'"'s face images.

12. The apparatus of claim 9, wherein, when a face image of a new candidate other than the existing candidates is input, the face feature extractor extracts a face feature used to register the new candidate by projecting the input face image on the base vector.

13. The apparatus of claim 9, further comprising a comparison search unit comparing a face feature used to confirm the identity of one candidate among existing candidates with face features of the existing candidates when a face image of the candidate is input and outputting a candidate list according to a result of the comparison,

wherein the face feature extractor extracts the face feature used to confirm the identity of the candidate by projecting the input face image on the base vector.

14. The apparatus of claim 12 or 13, wherein previous statistic values which are values calculated by each component of the apparatus are updated based on the previous statistic values and the input face image.