FACE RECOGNITION APPARATUS AND METHOD FOR CONTROLLING THE SAME

• US 20130108123A1
• Filed: 11/01/2012
• Published: 05/02/2013
• Est. Priority Date: 11/01/2011
• Status: Active Grant

• Current Assignee: Samsung Electronics Co. Ltd.
• Original Assignee: Samsung Electronics Co. Ltd.
• Inventors: Roh, Kyung Shik, Shim, Young Bo, Yoon, Suk June, HWANG, Won Jun
• Granted Patent: US 8,861,805 B2
• Time in Patent Office: Days
• US Class Current: 382/118
• CPC Class Codes: G06F 18/295 Markov models or related models, e.g. semi-Markov models; Markov random fields; Networks embedding Markov models
• CPC Class Codes: G06V 10/449 Biologically inspired filters, e.g. difference of Gaussians [DoG] or Gabor filters
• CPC Class Codes: G06V 10/85 Markov-related models; Markov random fields
• CPC Class Codes: G06V 40/168 Feature extraction; Face representation
• CPC Class Codes: G06V 40/172 Classification, e.g. identification

Abstract

A face recognition apparatus and face recognition method perform face recognition of a face by comparing an image of the face to be identified with target images for identification. The face recognition apparatus includes an image input unit to receive an image of a face to be identified, a sub-image production unit to produce a plurality of sub-images of the input face image using a plurality of different face models, a storage unit to store a plurality of target images, and a face recognition unit to set the sub-images to observed nodes of a Markov network, to set the target images to hidden nodes of the Markov network, and to recognize the presence of a target image corresponding to the face images to be identified using a first relationship between the observed nodes and the hidden nodes and a second relationship between the hidden nodes.

Claims

1. A face recognition apparatus comprising:

an image input unit to receive an image of a face to be identified;

a sub-image production unit to produce a plurality of sub-images of the input face image using a plurality of different face models; and

a face recognition unit to set the sub-images to observed nodes of a Markov network, to set target images to hidden nodes of the Markov network using the Markov network, and to recognize the presence of a target image corresponding to the face image to be identified using a first relationship between the observed nodes and the hidden nodes and a second relationship between the hidden nodes.

2. The face recognition apparatus according to claim 1, further comprising a storage unit to store the plurality of target images.

3. The face recognition apparatus according to claim 2, wherein the first relationship between the observed node and the hidden node is based on a similarity between the sub-image and the target image, and

the second relationship between the hidden nodes is based on a similarity between the target images.

4. The face recognition apparatus according to claim 3, wherein the face recognition unit comprises:

a first calculation unit to calculate a first similarity between the respective sub-images produced in the sub-image production unit and the target images stored in the storage unit and to array the target images in a predetermined order; and

a second calculation unit to calculate a second similarity between target images arrayed with respect to one sub-image among the sub-images, and target images arrayed with respect to at least one other sub-image among the sub-images.

6. The face recognition apparatus according to claim 4, wherein the second calculation unit calculates the second similarity from all sub-images produced by the sub-image production unit.

7. The face recognition apparatus according to claim 4, wherein the first calculation unit calculates the first relationship in accordance with the following Equation 2:

φ

_i(x_i,y_i)=exp(−

|s_i(x_i,y_i)−

1|²/2σ

²) 

 

[Equation 2]

wherein x_i and y_i are a target image i and a sub-image i, respectively, φ

(x_i, y_i) is a vector function representing a first relationship between the sub-image i and the target image i, s(x_i,y_i) represents a first similarity between the target image i and the sub-image i, and σ

represents a noise parameter.

8. The face recognition apparatus according to claim 4, wherein the second calculation unit calculates the second relationship in accordance with the following Equation 5:

ψ

_ij(x_i,x_j)=exp(−

|s_ij(x_i,x_j)−

1|²/2σ

²) 

 

[Equation 5]

wherein x_i and x_j are a target image i and a target image j, respectively, is a matrix function representing a second relationship between the target image i and the target image j, s(x_i, x_j) represents a second similarity between the target image i and the target image j, and σ

represents a noise parameter.

9. The face recognition apparatus according to claim 7, further comprising:

an identification unit to detect the target image corresponding to the face image to be identified by applying a belief propagation algorithm to the first relationship and the second relationship.

10. The face recognition apparatus according to claim 8, further comprising:

an identification unit to detect the target image corresponding to the face image to be identified by applying a belief propagation algorithm to the first relationship and the second relationship.

11. The face recognition apparatus according to claim 9, wherein the identification unit calculates a marginal probability based on the application result of the belief propagation algorithm and the first relationship, and detects the target image corresponding to the face image to be identified based on the calculated marginal probability.

5. The face recognition apparatus according to claim 5, wherein the target images are arrayed in an order based on a similarity from a high similarity to a relatively lower similarity.

12. A method for controlling a face recognition apparatus, the method comprising:

inputting a face image to be identified;

producing a plurality of sub-images of the input face image using a plurality of different face models;

setting the sub-images to observed nodes of a Markov network and setting target images to hidden nodes of the Markov network using the Markov network; and

recognizing a target image corresponding to the face image to be identified using a first relationship between the observed node and the hidden node and a second relationship between the hidden nodes.

13. The method according to claim 12, wherein the first relationship between the observed node and the hidden node is based on a similarity between the sub-image and the target image, and

the second relationship between the hidden nodes is based on a similarity between the target images.

14. The method according to claim 13, wherein the recognizing a target image corresponding to the face image to be identified comprises:

calculating a first similarity between the respective sub-images produced in the sub-image production unit and the target images and arraying the target images in a predetermined order; and

calculating a second similarity between target images arrayed with respect to one sub-image among the sub-images, and target images arrayed with respect to at least one other sub-image among the sub-images.

15. The method according to claim 14 wherein the second similarity is calculated based on all the produced sub-images.

16. The method according to claim 15, wherein the first relationship is calculated in accordance with the following Equation 2:

φ

_i(x_i,y_i)=exp(−

|s_i(x_i,y_i)−

1|²/2σ

²) 

 

[Equation 2]

wherein x_i and y_i are a target image i and a sub-image i, respectively, φ

(x_i,y_i) is a vector function representing a first relationship between the sub-image i and the target image i, s(x_i,y_i) represents a first similarity between the target image i and the sub-image i, and σ

represents a noise parameter.

17. The method according to claim 15, wherein the second relationship is calculated in accordance with the following Equation 5:

ψ

_ij(x_i,x_j)=exp(−

|s_ij(x_i,x_j)−

1|²/2σ

²) 

 

[Equation 5]

wherein x_i and x_i are a target image i and a target image j, respectively, φ

(x_i,x_j) is a matrix function representing a second relationship between the target image i and the target image j, s(x_i, x_j) represents a second similarity between the target image i and the target image j, and σ

represents a noise parameter.

18. The method according to claim 16, wherein the recognizing a target image corresponding to the face image to be identified is carried out by applying a belief propagation algorithm to the first relationship and the second relationship.

19. The method according to claim 17, wherein the recognizing a target image corresponding to the face image to be identified is carried out by applying a belief propagation algorithm to the first relationship and the second relationship.

20. The method according to claim 18, wherein the recognizing a target image corresponding to the face image to be identified is carried out by calculating a marginal probability based on the application result of the belief propagation algorithm and the first relationship, and recognizing the presence of a target image corresponding to the face image to be identified, based on the calculated marginal probability.

21. A face recognition apparatus comprising:

an image input unit to receive an image;

a sub-image production unit to produce a plurality of sub-images of the input image using a plurality of models;

a feature extraction unit to receive the plurality of sub-images generated by the sub-image production unit and to extract at least one feature from each of the plurality of sub-images; and

a recognition unit to set the sub-images to observed nodes of a Markov network, to set target images to hidden nodes of the Markov network using the Markov network, and to recognize the presence of a target image corresponding to the image to be identified by using the extracted features to determine a first relationship between the observed nodes and the hidden nodes and by determining a second relationship between the hidden nodes.

22. The face recognition apparatus according to claim 21, wherein the feature extraction unit comprises a plurality of extraction units which correspond to the plurality of sub-images, each extraction unit including at least one Gabor filter to obtain a plurality of features and a linear discriminant analyzer to perform linear discriminant analysis on the obtained features to calculate a feature vector.

23. The face recognition apparatus according to claim 22, wherein each Gabor filter obtains a plurality of features using an I-type extended curvature Gabor filter group using at least one parameter from among parameters including orientation, scale, Gaussian width, or curvature ratio.

24. The face recognition apparatus according to claim 23, wherein the I-type extended curvature Gabor filter group comprises a plurality of I-type curvature Gabor filters, wherein a sub-set of I-type curvature Gabor filters is selected from among the plurality of I-type curvature Gabor filters, and filtering is performed using the selected sub-set of I-type curvature Gabor filters.