Method and apparatus for extracting facial features from image containing face

• US 8,111,880 B2
• Filed: 08/30/2007
• Issued: 02/07/2012
• Est. Priority Date: 02/15/2007
• Status: Expired due to Fees

• Current Assignee: Samsung Electronics Co. Ltd.
• Original Assignee: Samsung Electronics Co. Ltd.
• Inventors: Lee, Jong-ha, Kim, Jung-bae, Moon, Young-su, Park, Gyu-tae, Hwang, Won-jun
• Primary Examiner(s): Werner, Brian
• Assistant Examiner(s): SHAH, UTPAL D
• Application Number:
  US11/896,300
• Publication Number:
  US 20080199055A1
• Time in Patent Office: 1,622 Days
• Field of Search: 382/118, 382/203, 382/260
• US Class Current: 382/118
• CPC Class Codes: G06V 10/449 Biologically inspired filters, e.g. difference of Gaussians [DoG] or Gabor filters
• CPC Class Codes: G06V 40/171 Local features and components; Facial parts eye characteristics G06V40/18; Occluding parts, e.g. glasses; Geometrical relationships

Abstract

Provided are a method and apparatus for extracting facial features from an image containing a face. The method and apparatus filter an input image using a filter set for face recognition at each of predetermined locations in the input image, merge values obtained by filtering the input image at locations, which are horizontally symmetrical to each other with respect to the center of the face, and synthesize values obtained by filtering the input image at locations which are not symmetrical to each other with respect to the center of the face with the merged values. Therefore, the time, feature values, and storage space required to extract or compare facial features can be significantly reduced. In addition, a face recognition system that runs well on low specification hardware can be implemented.

Claims

1. A method of extracting facial features from an image containing a face, the method comprising:

(a) filtering the image using a filter set for face recognition at each of predetermined locations in the image, the filtering including filtering one or more pairs of first values filtered at horizontally symmetrical locations from among the predetermined locations;

(b) merging each pair of the first values filtered at the horizontally symmetrical locations from among the predetermined locations, wherein the horizontally symmetrical locations are horizontally symmetrical to each other with respect to a longitudinal axis disposed along a center of the face, into second values representing the each pair of the first values at the horizontally symmetrical locations; and

(c) synthesizing third values filtered at locations, which are not horizontally symmetrical locations, from among the predetermined locations, with the second values, when the predetermined locations include locations other than the horizontally symmetrical locations.

2. The method of claim 1, wherein operation (a) comprises:

(a1) filtering the image using the filter set at the horizontally symmetrical locations from among the predetermined locations, which are horizontally symmetrical to each other with respect to the center of the face; and

(a2) filtering the image using the filter set at locations excluding the horizontally symmetrical locations from among the predetermined locations.

3. The method of claim 2, wherein the first values obtained by filtering the image in operation (a1) are merged in operation (b), and the third values obtained by filtering the image in operation (a2) are synthesized with the second values merged in operation (b) in operation (c).

4. The method of claim 1, wherein the filter set for face recognition comprises a predetermined number of filters which are classified according to scale, direction, and real number and imaginary number.

5. The method of claim 4, wherein operation (b) comprises merging the each pair of the first values using symmetric filters at two horizontally symmetrical locations which are horizontally symmetrical to each other with respect to the center of the face.

6. The method of claim 5, wherein operation (b) comprises averaging the each pair of the first values.

7. The method of claim 5, wherein operation (b) comprises assigning a weight to a value, which is obtained by filtering the image at a reliable location, among the each pair of the first values and merging the weighted value and the other value.

8. A non-transitory computer-readable recording medium on which a program for executing the method of claim 1 in a computer is recorded.

9. An apparatus for extracting facial features from an image containing a face, the apparatus comprising:

a filtering unit filtering the image using a filter set for face recognition at each of predetermined locations in the image, wherein the filtering unit filters one or more pairs of first values filtered at horizontally symmetrical locations from among the predetermined locations;

a merging unit merging, using at least one processor, each pair of the first values filtered at the horizontally symmetrical locations from among the predetermined locations, wherein the horizontally symmetrical locations are horizontally symmetrical to each other with respect to a longitudinal axis disposed along a center of the face, into second values representing the each pair of the first values at the horizontally symmetrical locations; and

a synthesis unit synthesizing third values filtered at locations, which are not horizontally symmetrical locations, from among the predetermined locations, with the second values, when the predetermined locations include locations other than the horizontally symmetrical locations.

10. The apparatus of claim 9, wherein the filtering unit comprises:

a first filtering unit filtering the image using the filter set at the horizontally symmetrical locations from among the predetermined locations, which are horizontally symmetrical to each other with respect to the center of the face; and

a second filtering unit filtering the image using the filter set at locations excluding the horizontally symmetrical locations from among the predetermined locations.

11. The apparatus of claim 10, wherein the merging unit merges the first values obtained by the first filtering unit, and the synthesis unit synthesizes the third values obtained by the second filtering unit with the second values merged by the merging unit.

12. The apparatus of claim 9, wherein the filter set for face recognition comprises a predetermined number of filters which are classified according to scale, direction, and real number and imaginary number.

13. The apparatus of claim 12, wherein the merging unit merges the each pair of the first values using symmetric filters at two horizontally symmetrical locations which are horizontally symmetrical to each other with respect to the center of the face.

14. The apparatus of claim 13, wherein the merging unit averages the each pair of the first values.

15. The apparatus of claim 13, wherein the merging unit assigns a weight to a value, which is obtained by filtering the image at a reliable location, from among the each pair of the first values and merges the weighted value and the other value.