Face identification apparatus, face identification method, and face identification program

• US 7,362,887 B2
• Filed: 10/18/2004
• Issued: 04/22/2008
• Est. Priority Date: 02/13/2004
• Status: Active Grant

• Current Assignee: Honda Motor Co., Ltd. (Honda Motor Company)
• Original Assignee: Honda Motor Co., Ltd. (Honda Motor Company)
• Inventors: Higaki, Nobuo, Sumida, Naoaki, Oohashi, Takahiro
• Primary Examiner(s): CHAWAN, SHEELA C
• Application Number:
  US10/965,919
• Publication Number:
  US 20050180611A1
• Time in Patent Office: 1,282 Days
• Field of Search: 382/117, 382/118, 382/168, 382/171, 382/172, 382/173, 382/190, 382/199, 382/203, 382/266, 382/274, 382/289, 382/296, 351/210, 351/206, 351/221, 351/204, 396/18, 348/78, 345/8
• US Class Current: 382/118
• CPC Class Codes: G06V 40/162 using pixel segmentation or colour matching
• CPC Class Codes: G06V 40/165 using facial parts and geometric relationships
• CPC Class Codes: G06V 40/167 using comparisons between temporally consecutive images

Abstract

A face identification apparatus for identifying a face is designed to have a face region expectation measure for expecting his/her face region within the image; a pupil-candidate-point detection measure for converting the face region to an image of a standard size, making it a standard image, and detecting his/her right/left pupil candidate points out of a search region within the standard image; a reference data generation measure for generating a normalization image from the standard image with making it a standard a distance of the right/left pupil candidate points and making reference data for evaluating advisability of the face region from the normalization image; and a face region evaluation measure for obtaining a degree of approximation between the reference data and standard data prepared in advance and evaluating the advisability of the face region.

Claims

1. A face identification apparatus comprising:

a face identification measure for identifying a face of an object person, based on an image of said object person taken by a camera;

a face region expectation measure for expecting a face region of said object person in said image, based on distance information representing a distance from said camera to said object person, and profile information and flesh-color region information of said object person generated from said image;

a pupil-candidate-point detection measure for generating a standard image with converting said face region to an image of a standard size, based on said distance information, setting a search region with making it a standard a center of gravity of said face region in the standard image, searching a circular edge for each pixel included in the search region, and detecting a right pupil candidate point and left pupil candidate point of said object person out of pixels of a predetermined circular edge;

a reference data generation measure for rotating said standard image so that a line segment connecting said right pupil candidate point and said left pupil candidate point becomes horizontal within said standard image, then generating a normalization image with making it a standard a distance of said right pupil candidate point and said left pupil candidate point, and making reference data for evaluating advisability of said face region from the normalization image; and

a face region evaluation measure for obtaining a degree of approximation between said reference data and standard data prepared in advance and evaluating the advisability of said face region, based on the obtained degree of approximation.

2. A face identification apparatus according to claim 1, wherein said pupil candidate point detection measure detects at least not less than each one of said right eye candidate point and said left eye candidate point, and said reference data generation measure generates said reference data for all combinations of said right eye candidate point and said left eye candidate point detected.

3. A face identification apparatus according to claim 2, wherein if said face region is judged inadequate by said face region evaluation measure, said face region evaluation measure requests said face region expectation to again expecting a face region of said object person.

4. A face identification apparatus according to claim 2, wherein if said face region is judged adequate by said face region evaluation measure, said face region evaluation measure makes said right eye candidate point and said left eye candidate point a standard and compensates said face region.

5. A face identification apparatus according to claim 1, wherein if said face region is judged inadequate by said face region evaluation measure, said face region evaluation measure requests said face region expectation to again expecting a face region of said object person.

6. A face identification apparatus according to claim 5, wherein if said face region is judged adequate by said face region evaluation measure, said face region evaluation measure makes said right eye candidate point and said left eye candidate point a standard and compensates said face region.

7. A face identification apparatus according to claim 1, wherein if said face region is judged adequate by said face region evaluation measure, said face region evaluation measure makes said right eye candidate point and said left eye candidate point a standard and compensates said face region.

8. A method for identifying a face of an object person, based on an image of said object person taken by a camera, the method comprising:

a face region expectation step of expecting the face region of said object person in said image, based on distance information representing a distance from said camera to said object person, profile information and flesh-color-region information of said object person generated from said image;

a pupil-candidate-point detection step of generating a standard image with converting said face region to an image of a standard size, based on said distance information, setting a search region with making it a standard a center of gravity of said face region in the standard image, searching a circular edge for each pixel included in the search region, and detecting a right pupil candidate point and left pupil candidate point of said object person out of pixels of a predetermined circular edge;

a reference data generation step of rotating said standard image so that a line segment connecting said right pupil candidate point and said left pupil candidate point becomes horizontal within said standard image, then generating a normalization image with making it a standard a distance of said right pupil candidate point and said left pupil candidate point, and making reference data for evaluating advisability of said face region from the normalization image;

a face region evaluation step of obtaining a degree of approximation between said reference data and standard data prepared in advance and evaluating the advisability of said face region, based on the obtained degree of approximation;

a face data making step of making face data of said object person from said normalization image if said face region is judged adequate in said face region evaluation step; and

a face identification step of comparing and collating the face data made in said face data making step with face data registered in a memory measure.

9. A face identification program, wherein in order to identify a face of an object person, based on an image of said object person taken by a camera, a computer is made to function as:

a face region expectation measure for expecting a face region of said object person in said image, based on distance information representing a distance from said camera to said object person, and profile information of and flesh-color-region information of said object person generated from said image;

a pupil-candidate-point detection measure for generating a standard image with converting said face region to an image of a standard size, based on said distance information, setting a search region with making it a standard a center of gravity of said face region in the standard image, searching a circular edge for each pixel included in the search region, and detecting a right pupil candidate point and left pupil candidate point of said object person out of pixels of a predetermined circular edge;

a reference data generation measure for rotating said standard image so that a line segment connecting said right pupil candidate point and said left pupil candidate point becomes horizontal within said standard image, then generating a normalization image with making it a standard a distance of said right pupil candidate point and said left pupil candidate point, and making reference data for evaluating advisability of said face region from the normalization image;

a face region evaluation measure for obtaining a degree of approximation between said reference data and standard data prepared in advance and expecting the advisability of said face region, based on the obtained degree of approximation;

a face data making step of making face data of said object person from said normalization image if said face region is judged adequate in said face region evaluation step; and

a face identification measure for comparing and collating the face data made in said face data making measure with face data registered in a memory measure.