Apparatus for identifying person

• US 5,995,639 A
• Filed: 11/29/1994
• Issued: 11/30/1999
• Est. Priority Date: 03/29/1993
• Status: Expired due to Fees

• Current Assignee: Matsushita Electric Industrial Company Limited (Panasonic Holdings Corporation)
• Original Assignee: Matsushita Electric Industrial Company Limited (Panasonic Holdings Corporation)
• Inventors: Nobori, Kunio, Kado, Yoshiyasu, Nakagawa, Masamichi
• Primary Examiner(s): Chang, Jon
• Assistant Examiner(s): Patel, Jayanti K.
• Application Number:
  US08/343,543
• Time in Patent Office: 1,827 Days
• Field of Search: 382/118, 382/115, 382/215, 382/203, 382/241, 382/190, 382/195, 382/201, 395/119, 395/120, 395/126, 395/127, 395/125, 395/152, 395/135, 395/118, 345/418-435, 345/473, 345/126, 345/138, 351/200
• US Class Current: 382/118
• CPC Class Codes: G06V 40/166 using acquisition arrangements
• CPC Class Codes: G06V 40/168 Feature extraction; Face representation
• CPC Class Codes: G06V 40/171 Local features and components; Facial parts eye characteristics G06V40/18; Occluding parts, e.g. glasses; Geometrical relationships

Abstract

The present invention inputs a face image through an image input section, extracts feature points of face components such as an eye, nose, mouth from the input face image, and divides the face image into many patches according to predetermined rules to describe the face image as a set of patches. Then the present invention extracts a feature amount for each patch to describe the face image as a set of feature amounts to identify a person.

Claims

1. An apparatus for identifying persons comprising:

an image input section that inputs a personal face image of a person to be identified;

an image memory section that stores the personal face image input by said image input section;

a feature points extracting section that extracts predefined feature points from the personal face image stored in said image memory section;

a feature amounts extracting section that divides the personal face image into small polygons based on the extracted feature points according to a predefined rule, thereby describing the personal face image as a set of the small polygons and extracts feature amounts for respective small polygons; and

a judgement section that calculates distances between the feature amounts extracted by said amounts extracting section and feature amounts extracted beforehand with respect to each person to be identified and judges whether the person having said personal face image input by said image input section is the person to be identified;

wherein said feature amounts extracting section comprises

a standard structure model that describes a presumed standard face as a set of small polygons defined on the basis of standard feature points predefined with respect to said standard face; and

a structure model adjustment section that deforms the standard structure model based on the feature points extracted by said points extracting section so as to match the personal face image with the deformed standard structure model,

and uses each difference in an average brightness between adjacent two small polygons of the matched personal face image as the feature amount.

2. An apparatus for identifying persons comprising:

an image input section that inputs a personal face image of a person to be identified;

an image memory section that stores the personal face image input by said image input section;

a feature points extracting section that extracts predefined feature points form the personal face image stored in said image memory section;

a feature amounts extracting section that divides the personal face image into small polygons based on the extracted feature points according to a predefined rule, thereby describing the personal face image as a set of the small polygons and extracts feature amounts for respective small polygons; and

a judgement section that calculates distances between the feature amounts extracted by said amounts extracting section and feature amounts extracted beforehand with respect to each person to be identified and judges whether the person having said personal face image input by said image input section is the person to be identified;

wherein average brightness at respective small polygons of the matched personal face image are used as the feature amounts, a position of a light source for illuminating a personal face relative to the personal face is estimated based on a distribution of the average brightness, and each average brightness is corrected based on the estimated position of the light source.

3. An apparatus for identifying persons comprising:

an image input section that inputs a personal face image of a personal to be identified;

an image memory section that stores the personal face image input by said image input section;

a feature points extracting section that extracts predefined feature points from the personal face image stored in said image memory section;

a feature amounts extracting section that divides the personal face image into small polygons based on the extracted feature points according to a predefined rule, thereby describing the personal face image as a set of the small polygons and extracts feature amounts for respective small polygons; and

a judgement section that calculates distances between the feature amounts extracted by said amounts extracting section and feature amounts extracted beforehand with respect to each person to be identified and judges whether the person having said personal face image input by said image input section is the person to be identified,

wherein said feature amounts extracting section comprises

a standard structure model that describes a presumed standard face as a set of small polygons defined on the basis of standard feature points predefined with respect to said standard face; and

a structure model adjustment section that deforms the standard structure model based on the feature points extracted by said points extracting section so as to match the personal face image with the deformed standard structure model,

wherein said standard structure model is a three-dimensional model defined by small polygons and their normal vectors,

wherein the normal vector of each small polygon is detected together with an average brightness of each small polygon, and

wherein a position of a light source for illuminating a personal face relative to the personal face is estimated based on the detected normal vectors and average brightness.