IMAGE PROCESSING APPARATUS AND METHOD, AND PROGRAM

• US 20090196467A1
• Filed: 01/21/2009
• Published: 08/06/2009
• Est. Priority Date: 02/05/2008
• Status: Active Grant

• Current Assignee: Sony Corporation (Sony Group Corp.)
• Original Assignee: Sony Corporation (Sony Group Corp.)
• Inventors: OKUBO, Atsushi
• Granted Patent: US 8,254,645 B2
• Time in Patent Office: Days
• US Class Current: 382/118
• CPC Class Codes: G06F 18/211 Selection of the most significant subset of features
• CPC Class Codes: G06V 10/772 Determining representative reference patterns, e.g. averaging or distorting patterns; Generating dictionaries
• CPC Class Codes: G06V 40/172 Classification, e.g. identification
• CPC Class Codes: G06V 40/50 Maintenance of biometric data or enrolment thereof

Abstract

An image processing apparatus includes a holding unit configured to hold, for each combination of a first angle indicating a face direction of a first face image which includes a human face and a second angle indicating a face direction of a second face image which includes a human face, a learning dictionary including information related to positions of feature points associating the first and second face images when a similarity degree between the first and second face images is estimated, a selection unit configured to select the learning dictionary held for each combination in accordance with the combination of the first and second angles, and a similarity degree estimation unit configured to estimate a facial similarity degree between the first and second face images on the basis of feature amounts extracted from the face images corresponding to the positions of the feature points included in the selected learning dictionary.

Claims

1. An image processing apparatus comprising:

holding means for holding, for each combination of a first angle indicating a direction of a face of a first face image which includes a human face and a second angle indicating a direction of a face of a second face image which includes a human face, a learning dictionary including information related to positions of feature points which associate the first face image with the second face image if a similarity degree between the first face image and the second face image is estimated;

selection means for selecting the learning dictionary held for each combination of the first angle and the second angle in accordance with the combination of the first angle and the second angle; and

similarity degree estimation means for estimating a facial similarity degree between the first face image and the second face image on the basis of feature amounts extracted from the first face image and the second face image corresponding to the positions of the feature points included in the selected learning dictionary.

2. The image processing apparatus according to claim 1, further comprising:

identification means for identifying the first face image and the second face image on the basis of the estimated similarity degree.

3. The image processing apparatus according to claim 1, wherein the second face image is registered in advance while being associated with the extracted feature amounts and the second angle.

4. The image processing apparatus according to claim 3, further comprising:

detection means for detecting a part of the first face image from an image including a human being;

facial feature detection means for detecting a facial feature position which is a feature of the face from the detected first face image;

face angle estimation means for estimating the first angle on the basis of the detected first face image and the facial feature position; and

extraction means for extracting a feature amount from an image in the vicinity of the facial feature position in the detected first face image,

wherein the select means selects the learning dictionary in accordance with the combination of the estimated first angle and the registered second angle, and

wherein the similarity degree estimation means estimates the similarity degree on the basis of the feature amount extracted from the first face image and the feature amount of the registered second face image corresponding to the positions of the feature points included in the selected learning dictionary.

5. The image processing apparatus according to claim 1, wherein the holding means holds the learning dictionaries for all the combinations in a case where the direction indicated by the first angle is set in three stages including a left direction, a front direction, and a right direction and the direction indicated by the second angle is set in three stages including a left direction, a front direction, and a right direction.

6. The image processing apparatus according to claim 2,

wherein the learning dictionary includes a predetermined threshold which functions as a reference for identifying the first face image and the second face image, and

wherein the identification means identifies whether the face of the first face image and the face the second face image are a same person or different persons on the basis of a result of a comparison between the estimated similarity degree and the threshold.

7. An image processing method comprising the steps of:

selecting a learning dictionary held for each combination of a first angle indicating a direction of a face of a first face image which includes a human face and a second angle indicating a direction of a face of a second face image which includes a human face, for each combination of the first angle and the second angle in accordance with the combination of the first angle and the second angle, the learning dictionary including information related to positions of feature points which associate the first face image with the second face image when a similarity degree between the first face image and the second face image is estimated; and

estimating a facial similarity degree between the first face image and the second face image on the basis of feature amounts extracted from the first face image and the second face image corresponding to the positions of the feature points included in the selected learning dictionary.

8. A program for instructing a computer to execute a processing which comprises the steps of:

selecting a learning dictionary held for each combination of a first angle indicating a direction of a face of a first face image which includes a human face and a second angle indicating a direction of a face of a second face image which includes a human face, for each combination of the first angle and the second angle in accordance with the combination of the first angle and the second angle, the learning dictionary including information related to positions of feature points which associate the first face image with the second face image when a similarity degree between the first face image and the second face image is estimated; and

estimating a facial similarity degree between the first face image and the second face image on the basis of feature amounts extracted from the first face image and the second face image corresponding to the positions of the feature points included in the selected learning dictionary.

9. An image processing apparatus comprising:

holding means for holding, for each combination of a first angle indicating a direction of an object or a living matter having a predetermined shape and feature of a first image which includes the object or the living matter and a second angle indicating a direction of an object or a living matter having a predetermined shape and feature of a second image which includes the object or the living matter, a learning dictionary including information related to positions of feature points which associate the first image with the second image when a similarity degree between the first image and the second image is estimated;

selection means for selecting the learning dictionary held for each combination of the first angle and the second angle in accordance with the combination of the first angle and the second angle; and

similarity degree estimation means for estimating a similarity degree of the object or the living matter between the first image and the second image on the basis of feature amounts extracted from the first image and the second image corresponding to the positions of the feature points included in the selected learning dictionary.

10. An image processing method comprising the steps of:

selecting a learning dictionary held for each combination of a first angle indicating a direction of an object or a living matter having a predetermined shape and feature of a first image which includes the object or the living matter and a second angle indicating a direction of an object or a living matter having a predetermined shape and feature of a second image which includes the object or the living matter, for each combination of the first angle and the second angle in accordance with the combination of the first angle and the second angle, the learning dictionary including information related to positions of feature points which associate the first image with the second image when a similarity degree between the first image and the second image is estimated; and

estimating a similarity degree of the object or the living matter between the first image and the second image on the basis of feature amounts extracted from the first image and the second image corresponding to the positions of the feature points included in the selected learning dictionary.

11. A program for instructing a computer to execute a processing which comprises the steps of:

selecting a learning dictionary held for each combination of a first angle indicating a direction of an object or a living matter having a predetermined shape and feature of a first image which includes the object or the living matter and a second angle indicating a direction of an object or a living matter having a predetermined shape and feature of a second image which includes the object or the living matter, for each combination of the first angle and the second angle in accordance with the combination of the first angle and the second angle, the learning dictionary including information related to positions of feature points which associate the first image with the second image when a similarity degree between the first image and the second image is estimated; and

estimating a similarity degree of the object or the living matter between the first image and the second image on the basis of feature amounts extracted from the first image and the second image corresponding to the positions of the feature points included in the selected learning dictionary.

12. An image processing apparatus comprising:

a holding unit configured to hold, for each combination of a first angle indicating a direction of a face of a first face image which includes a human face and a second angle indicating a direction of a face of a second face image which includes a human face, a learning dictionary including information related to positions of feature points which associate the first face image with the second face image when a similarity degree between the first face image and the second face image is estimated;

a selection unit configured to select the learning dictionary held for each combination of the first angle and the second angle in accordance with the combination of the first angle and the second angle; and

a similarity degree estimation unit configured to estimate a facial similarity degree between the first face image and the second face image on the basis of feature amounts extracted from the first face image and the second face image corresponding to the positions of the feature points included in the selected learning dictionary.

13. An image processing apparatus comprising:

a holding unit configured to hold, for each combination of a first angle indicating a direction of an object or a living matter having a predetermined shape and feature of a first image which includes the object or the living matter and a second angle indicating a direction of an object or a living matter having a predetermined shape and feature of a second image which includes the object or the living matter, a learning dictionary including information related to positions of feature points which associate the first image with the second image when a similarity degree between the first image and the second image is estimated;

a selection unit configured to select the learning dictionary held for each combination of the first angle and the second angle in accordance with the combination of the first angle and the second angle; and

a similarity degree estimation unit configured to estimate a similarity degree of the object or the living matter between the first image and the second image on the basis of feature amounts extracted from the first image and the second image corresponding to the positions of the feature points included in the selected learning dictionary.