Method for verifying users and updating database, and face verification system using the same

• US 7,187,786 B2
• Filed: 04/22/2003
• Issued: 03/06/2007
• Est. Priority Date: 04/23/2002
• Status: Active Grant

• Current Assignee: Samsung Electronics Co. Ltd.
• Original Assignee: Samsung Electronics Co. Ltd.
• Inventors: Kee, Seok-cheol
• Primary Examiner(s): CHAWAN, SHEELA C
• Application Number:
  US10/419,742
• Publication Number:
  US 20030198368A1
• Time in Patent Office: 1,414 Days
• Field of Search: 382/118, 382/124, 382/127, 382/159, 382/190, 382/225, 382/170, 382/226, 382/227, 382/245, 382/296, 382/224, 378/98.2, 704/232, 704/273, 704/246, 704/231, 713/186, 706/20
• US Class Current: 382/118
• CPC Class Codes: G06V 40/16 Human faces, e.g. facial parts, sketches or expressions
• CPC Class Codes: G07C 9/37 using biometric data, e.g. fingerprints, iris scans or voice recognition

Abstract

To reduce degradation of recognition performance due to eye detection errors during face verification and to overcome a problem in that sufficient data to design an optimum feature classifier cannot be obtained during face registration, a method includes shifting the positions of eyes detected during face registration in predetermined directions by a predetermined distance to generate pairs of new coordinate points of the eyes; normalizing a face image on the basis of each pair of new coordinate points of the eyes; using the results of normalization in teaching a feature classifier, thereby coping with eye detection errors. In addition, two threshold values are used to prevent a database from being updated with a face of an unregistered person and to update the database with a normal client'"'"'s face image that has been used during the latest face verification.

Claims

1. A method of verifying a face in a face verification system, comprising the steps of:

(a) separating background and a face from an input face image to detect a face region;

(b) detecting the positions of eyes from the detected face region and shifting the detected positions of eyes in predetermined directions by a predetermined distance, thereby generating new coordinate points of the eyes;

(c) normalizing the face region on the basis of the new coordinate points of the eyes;

(d) extracting recognition features from the normalized face regions and calculating feature values;

(e) teaching a feature classifier using the feature values and storing the feature values in a database;

(f) detecting a face region from another input face image for verification, normalizing the face region, extracting recognition features of a face, and calculating feature values; and

(g) determining whether the input face image is similar to any face image registered in the database based on the feature values.

2. The method of claim 1, wherein step (g) comprises calculating the distribution of the calculated feature values in a vector area using the feature classifier and determining whether the input face image is similar to any face image registered in the database based on a comparison value which numerically expresses the distribution of the calculated feature values.

3. The method of claim 2, wherein step (g) further comprises rejecting the user if the comparison value is equal to or less than a first threshold value and accepting the user if the comparison value is greater than the first threshold value.

4. The method of claim 3, further comprising updating the database with the face image for verification by performing the steps (b) through (d) on the face image for verification when the comparison value is greater than a second threshold value.

5. The method of claim 4, wherein the second threshold value is set to be greater than the first threshold value such that a probability of a user'"'"'s passing through the face verification system, the user'"'"'s face image not being stored in the database, is zero, and as the comparison value increases, a probability of the face image for verification agreeing with a face image previously stored in the database increases.

6. The method of claim 4, wherein the second threshold value is determined by ƒ

(x)=sgn(<

w*x>

+b), wherein sgn indicates a hyperplane function, w indicates the slope of a hyperplane that is obtained through learning, x indicates a feature value input during face verification, and the threshold value is determined by adjusting a bias value b.

7. The method of claim 3, wherein the first threshold A value is determined by ƒ

(x)=sgn(<

w*x>

+b), wherein sgn indicates a hyperplane function, w indicates the slope of a hyperplane that is obtained through learning, x indicates a feature value input during face verification, and the threshold value is determined by adjusting a bias value b.

8. The method of claim 1, further comprising converting the input image into a format suitable for image processing before the step (a).

9. The method of claim 1, wherein the step (c) comprises calculating the center point between each pair of new coordinate points of the eyes and normalizing the face region by compensating for a tilt of the face region on the basis of each center point.

10. The method of claim 1, wherein in step (b) the predetermined distance is determined according to a predetermined ratio of the predetermined distance to a horizontal length of the detected face region.

11. The method of claim 10, wherein the predetermined ratio is 1/15.

12. The method of claim 1, where in step (b) the predetermined directions are four directions, i.e., up, down, to the left, or to the right.

13. A computer readable medium storing a program which carries out the method of claim 1.

14. A face verification system comprising:

a face region detector for separating background and a face from an input face image to detect a face region;

an eye-position shift unit for detecting the positions of eyes from the detected face region and shifting the detected positions of eyes in predetermined directions by a predetermined distance, thereby generating new coordinate points of the eyes;

a face region normalizer for normalizing the face region on the basis of the new coordinate points of the eyes;

a recognition feature extractor for extracting recognition features from the normalized face regions and calculating feature values;

a feature classifier teacher for teaching a feature classifier using the calculated feature values;

a database for storing the calculated feature values; and

a determiner for determining whether another input face image is similar to any face image registered in the database based on the feature values.

15. The face verification system of claim 14, further comprising a preprocessor for processing the input face image to have a format suitable for image processing and outputting the result of the processing to the face region detector.

16. The face verification system of claim 15, wherein the predetermined ratio is 1/15.

17. The face verification system of claim 14, wherein the eye-position shift unit shifts the detected positions of the eyes in four directions, i.e., up, down, to the left, and to the right, by the predetermined distance in accordance with a predetermined ratio of the predetermined distance to a horizontal length of the detected face region, thereby generating pairs of the new coordinate points of the eyes.

18. The face verification system of claim 14, wherein the determiner calculates a comparison value which numerically expresses the distribution of the feature values, which are extracted from the face image for verification, in a vector area, and compares the comparison value with threshold values, which are set according to the distribution of feature values of previously registered users, to determine whether the input face image is similar to one of the registered face images and whether to update the database.

19. The face verification system of claim 18, wherein the threshold values are determined by ƒ

(x)=sgn(<

w*x>

+b), wherein sgn indicates a hyperplane function, w indicates the slope of a hyperplane that is obtained through learning, x indicates a feature value input during face verification, and the threshold values are determined by adjusting a bias value b.

20. The face verification system of claim 14, wherein the determiner further determines whether to update the database using the input face image.