Ethnicity Classification Using Multiple Features

• US 20110249891A1
• Filed: 04/07/2010
• Published: 10/13/2011
• Est. Priority Date: 04/07/2010
• Status: Active Grant

• Current Assignee: 138 East LCD Advancements Limited (Vector Capital Corporation)
• Original Assignee: Seiko Epson Corporation (Seiko Group)
• Inventors: Li, Jia
• Granted Patent: US 8,331,698 B2
• Time in Patent Office: Days
• US Class Current: 382/165
• CPC Class Codes: G06F 18/254 of classification results, e.g. of results related to same input data
• CPC Class Codes: G06V 10/809 of classification results, e.g. where the classifiers operate on the same input data
• CPC Class Codes: G06V 40/171 Local features and components; Facial parts eye characteristics G06V40/18; Occluding parts, e.g. glasses; Geometrical relationships
• CPC Class Codes: G06V 40/172 Classification, e.g. identification

Abstract

Frontal face images are classified into four categories such as Asian, Caucasian, African and others. A new representation of face appearance, named BITF (Block Intensity and Texture Feature), is employed as the discrimination feature. An ensemble of three component classifiers, each trained with a different number of BITF features as inputs, is designed to achieve a reliable classification. Further reliability is obtained by taking into consideration other secondary features to boost the classification performance.

Claims

1. A method for automatic ethnicity classification of a frontal face image comprising:

aligning the frontal face image to a template;

normalizing the aligned frontal face image;

generating a block intensity and texture feature (BITF) from the normalized frontal face image; and

using a BITF classifier ensemble to classify the frontal face image into one of four ethnicity classifications.

2. A method as recited in claim 1 wherein generating a block intensity and texture feature (BITF) comprises:

dividing the normalized frontal face image into non-overlapping blocks of pixels;

calculating a mean intensity of each block of pixels;

calculating a co-occurrence matrix of each block of pixels to describe texture; and

concatenating the mean intensities and co-occurrence matrices of all blocks to generate one feature vector.

3. A method as recited in claim 1, further comprising:

deciding if using the BITF classifier ensemble provides a confident discriminate result, and if the BITF classifier ensemble does not provide a confident discriminate result, then using a secondary classifier to classify a secondary feature.

4. A method as recited in claim 3, wherein the secondary feature comprises skin tone.

5. A method as recited in claim 3, wherein the secondary classifier is a radial basis function (RBF) classifier.

6. A method as recited in claim 3, further comprising:

combining classification results of the BITF classifier ensemble and the secondary classifier to classify the frontal face image.

7. A method as recited in claim 6, further comprising:

deciding if combining the classification results of the BITF classifier ensemble and the secondary classifier provides a confident discriminate result, and if combining the classification results of the BITF classifier ensemble and the secondary classifier does not provide a confident discriminate result, then using the secondary classifier to classify another secondary feature.

8. A method as recited in claim 7, further comprising:

combining classification results of the BITF classifier ensemble, the secondary classifier classification using the secondary feature, and the secondary classifier classification using the another secondary feature to classify the frontal face image.

9. A method as recited in claim 7, wherein the another secondary feature comprises hair color, eye color, or eyebrow to eye distance.

10. A method as recited in claim 1 wherein the four ethnicity classifications comprise Asian, Caucasian, African, and others.

11. One or more computer-readable media having computer-readable instructions thereon, which, when executed by a processor, implement a method for automatically classifying an ethnicity of a front facial image, comprising:

aligning the frontal face image to a template;

normalizing the aligned frontal face image;

generating a block intensity and texture feature (BITF) from the normalized frontal face image; and

using a BITF classifier ensemble to classify the frontal face image into one of four ethnicity classifications.

12. The one or more computer-readable media as recited in claim 11, wherein generating a block intensity and texture feature (BITF) comprises:

dividing the normalized frontal face image into non-overlapping blocks of pixels;

calculating a mean intensity of each block of pixels;

calculating a co-occurrence matrix of each block of pixels to describe texture; and

concatenating the mean intensities and co-occurrence matrices of all blocks to generate one feature vector.

13. The one or more computer-readable media as recited in claim 11, further comprising:

deciding if using the BITF classifier ensemble provides a confident discriminate result, and if the BITF classifier ensemble does not provide a confident discriminate result, then using a secondary classifier to classify a secondary feature.

14. The one or more computer-readable media as recited in claim 13, wherein the secondary classifier is a radial basis function (RBF) classifier.

15. The one or more computer-readable media as recited in claim 13, further comprising:

combining classification results of the BITF classifier ensemble and the secondary classifier to classify the frontal face image.

16. A device for automatic ethnicity classification of a frontal face image comprising:

a processor that;

aligns the frontal face image to a template;

normalizes the aligned frontal face image;

generates a block intensity and texture feature (BITF) from the normalized frontal face image; and

uses a BITF classifier ensemble to classify the frontal face image into one of four ethnicity classifications.

17. The device as recited in claim 16, wherein generating a block intensity and texture feature (BITF) comprises:

dividing the normalized frontal face image into non-overlapping blocks of pixels;

calculating a mean intensity of each block of pixels;

calculating a co-occurrence matrix of each block of pixels to describe texture; and

concatenating the mean intensities and co-occurrence matrices of all blocks to generate one feature vector.

18. The device as recited in claim 16, wherein the processor further:

decides if using the BITF classifier ensemble provides a confident discriminate result, and if the BITF classifier ensemble does not provide a confident discriminate result then uses a secondary classifier to classify a secondary feature.

19. The device as recited in claim 18, wherein the secondary classifier is a radial basis function (RBF) classifier.

20. The device as recited in claim 18, wherein the processor further:

combines classification results of the BITF classifier ensemble and the secondary classifier to classify the frontal face image.