The present invention claims priority of Korean Patent Application No. 10-2008-0117354, filed on Nov. 25, 2008, which is incorporated herein by reference.
The present invention relates to a method and apparatus for detecting a forged face using an infrared image, and, more particularly, to a method and apparatus for detecting a forged face using an infrared image in which a face is imaged by simultaneously using an infrared camera and a still camera, and analyze the characteristics of an input infrared image to determine whether a current face is a forged face.
As is well known in the art, biometrics technology is the general term for methods of authenticating individual persons based on physiological or behavioral characteristics. The biometrics technology is characterized in that there is no possibility of losing biological information and there is no need to memorize such biological information. Further, biometrics technology is advantageous in that, since the biological information of an individual needs to be personally input to a system, biometrics technology is more secure than existing technology based on a password from the standpoint of security, and thus has attracted attention as the next-generation technology which will replace existing personal authentication technology based on a password and Identification (ID).
However, as the conventional biometric recognition technology progresses, a forged biological information generation technology is also developed, so that it is difficult to distinguish actual biological information from the forged biological information using typical biometric recognition algorithm. Accordingly, the reliability and security of biometric recognition are deteriorated.
Further, in conventional face recognition, when thermal distribution of a face captured by a thermal infrared camera is used, a forged face and an actual face can be easily distinguished from each other. However, since equipment such as a thermal infrared camera is very expensive, there is a limitation in actually applying such equipment to a system. Furthermore, a method of forcing a user to speak or move has a problem of inconveniencing a user.
Therefore, the present invention provides a method and apparatus for detecting a forged face using an infrared image which capture face images by simultaneously using an infrared camera and a still camera, determine whether a current face is a forged face by analyzing the characteristics of an input infrared image, and perform user authentication using a photorealistic image captured by visible light if it has been determined that the current face is not a forged face.
In accordance with an aspect of the present invention, there is provided a method for detecting a forged face using an infrared image, including: acquiring an infrared image and a photorealistic image captured by one camera or two cameras; extracting a face region from the photorealistic image; determining, based on analysis of quality characteristics of the infrared image, whether a current face is a forged face or not; and performing face recognition on the extracted face region if it is determined that the current face is a non-forged face.
In accordance with another aspect of the present invention, there is provided an apparatus for detecting a forged face using an infrared image, including: an image acquisition unit for acquiring an infrared image and a photorealistic image captured by one or two cameras; a face region extraction unit for extracting a face region from the photorealistic image; a forged face determination unit for determining, based on analysis of quality characteristics of the infrared image, whether a current face is a forged face; and a face recognition unit for performing face recognition on the extracted face region if it is determined that the current face is a non-forged face.
Accordingly, the present invention captures face images by simultaneously using an infrared camera and a still camera, determines whether a current face is a forged face by analyzing the characteristics of an input infrared image, and performs user authentication using a photorealistic image captured by visible light if it is determined that the current face is not a forged face, thus solving a plurality of problems of existing face recognition methods in which it is difficult to distinguish actual face information from forged face information, there is a limitation due to the use of expensive equipment, and a user is inconvenienced.
Furthermore, the present invention is advantageous in that it captures face images by simultaneously using an infrared camera and a still camera, determines whether a current face is a forged face by analyzing the characteristics of an input infrared image, and performs user authentication using a photorealistic image captured by visible light if it is determined that the current face is not a forged face, thus improving the reliability and security of a face detection device.
The objects and features of the present invention will become apparent from the following description of preferred embodiments given in conjunction with the accompanying drawings, in which:
Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings which form a part hereof.
The image acquisition unit 10 includes an infrared image acquisition unit 11 and a photorealistic image acquisition unit 13.
The infrared image acquisition unit 11 acquires an infrared image captured by an infrared camera S1 to provide it to the forged face determination unit 30.
The photorealistic image acquisition unit 13 acquires a photorealistic image captured by a still camera S2 to provide it to the face region extraction unit 20.
The face region extraction unit 20 extracts a face region from the photorealistic image input from the photorealistic image acquisition unit 13 by applying a component-based image processing technique, or pattern recognition technique using Adaboost or a support vector machine (hereinafter referred to as an ‘SVM’) to the photorealistic image, and provides extracted face region information to the forged face determination unit 30.
As shown in the detailed block diagram in
The face image evaluation unit 31 evaluates the quality, i.e., clearness, of a region in the infrared image by using sharpness technique, the region of the infrared image corresponding to the face region extracted and input from the face region extraction unit 20. When, as a result of the evaluation, a clearness degree of the infrared image region is less than a preset clearness value, the face image evaluation unit 31 determines a current face to be a forged face, and provides the determination result to the face recognition unit 40. In contrast, when, as a result of the evaluation, the clearness degree of the face region satisfies the preset clearness value, the face image evaluation unit 31 determines the current face to be a non-forged face, and provides the determination result to the face recognition unit 40.
The face presence detection unit 33 detects, based on characteristics of the infrared image indicating that a facial shape is not visualized in a case of the infrared image captured from the forged face, whether a facial shape is present or not by applying the pattern recognition technology to the face region. When, as a detection result, a facial shape is not present, the face presence detection unit 33 determines the current face to be a forged face, and provides the determination result to the face recognition unit 40. In contrast, when, as a result of the detection, a facial shape is present, the face presence detection unit 33 determines the current face to be a non-forged face, and provides the determination result to the face recognition unit 40.
The eyeball characteristic analysis unit 35 compares preset eyeball regions and an eyeball region of the infrared image input from the infrared image acquisition unit 11, the preset eyeball region being provided by detecting eyeball regions from actual infrared images of various users in advance. When, as a result of the comparison, the eyeball region of the input infrared image is not identical to any of the preset eyeball regions, the eyeball characteristic analysis unit 35 determines the current face to be a forged face, and provides the determination result to the face recognition unit 40. In contrast, when, as a result of the analysis, the eyeball region of the infrared image is identical to one of the preset eyeball regions, the eyeball characteristic analysis unit 35 determines the current face to be a non-forged face, and provides the determination result to the face recognition unit 40.
The face recognition unit 40 performs face recognition on the face region input from the face region extraction unit 20, depending on the determination results input from the forged face determination unit 30 when the determination results indicate that the current face is a non-forged face. In contrast, the face recognition unit 40 does not perform face recognition and may perform control such that the current face again is captured again, when the determination results input from the forged face determination unit 30 indicate that the current face is a forged face.
Accordingly, the present invention captures face images by simultaneously using an infrared camera and a still camera, determines whether a current face is a forged face by analyzing the characteristics of an input infrared image, and performs user authentication using a photorealistic image captured by visible light if it is determined that the current face is a non-forged face, thus solving a plurality of problems in the prior art in which it is difficult to distinguish actual face information from forged face information using existing face recognition methods, there is a limitation due to the use of expensive equipment, and a user is inconvenienced.
Hereinafter, a method for detecting a forged face by using the above described apparatus for detecting a forged face in accordance with the present embodiment will be described below.
First, an infrared image and a photorealistic image are respectively captured by the infrared camera S1 and the still camera S2 at step S301. Here, the infrared image and photorealistic image may also be captured by one camera which is configured to obtain both images by using visible light and infrared light.
Then, the infrared image acquisition unit 11 in the image acquisition unit 10 acquires the infrared image from the infrared camera S1 to provide the infrared image to the forged face determination unit 30 at step S303. The photorealistic image acquisition unit 13 acquires the photorealistic image from the still camera S2 to provide the photorealistic image to the face region extraction unit 20 at step S305.
Next, the face region extraction unit 20 extracts a face region from the photorealistic image input from the photorealistic image acquisition unit 13 by applying a component-based image processing technique, or pattern recognition technology using Adaboost or SVM, to the photorealistic image at step S307, and provides the extracted face region to the forged face determination unit 30.
Then, the face image evaluation unit 31 in the forged face determination unit 30 evaluates the quality of the region in the infrared image which corresponds to the face region extracted by and input from the face region extraction unit 20 by using sharpness technique at step S309. That is, at step S309, if the clearness degree of the infrared image region is less than a preset clearness value, the face image evaluation unit 31 determines the current face to be a forged face and provides the determination result to the face recognition unit 40. If the clearness of the infrared image region is equal to or greater than the preset clearness value, the face image evaluation unit 31 determines the current face to be a non-forged face and provides the determination result to the face recognition unit 40.
For example,
Next, the face presence detection unit 33 of the forged face determination unit 30 detects whether a facial shape is present or not by applying Adaboost or SVM technology to the face region extracted by and input from the face region extraction unit 20, on the basis of the characteristics of the infrared image input from the infrared image acquisition unit 11 at step S311. That is, at step S311, if a facial shape is not present, the face presence detection unit 33 determines the current face to be a forged face and provides the determination result to the face recognition unit 40. If a facial shape is present, the face presence detection unit 33 determines the current face to be a non-forged face, and provides the determination result of the determination to the face recognition unit 40.
In this case, in order to promptly detect the presence/non-presence of a facial shape, a region for detection is limited to the face region extracted by and input from the face region extraction unit 20, but, if necessary, the presence/non-presence of a facial shape may be detected by applying a face detection technique to the entire infrared image.
Next, at step S313, the eyeball characteristic analysis unit 35 in the forged face determination unit 30 compares preset eyeball regions and an eyeball region of the infrared image input from the infrared image acquisition unit 11, the preset eyeball region being provided by detecting eyeball regions from actual infrared images of various users in advance.
As a result of the analysis at step S313, if the eyeball region of the input infrared image does not have characteristics identical to one of the preset eyeball regions, the eyeball characteristic analysis unit 35 determines the current face to be a forged face, and provides the determination result to the face recognition unit 40. If the eyeball region of the input infrared image has characteristics identical to one of the preset eyeball regions, the eyeball characteristic analysis unit 35 determines the current face not to be a forged face and provides the determination results to the face recognition unit 40.
At step S315, if the determination results indicate that the current face is not a forged face, the face recognition unit 40 performs face recognition on the face region input from the face region extraction unit 20 at step S317. Otherwise, the current face is captured again by the infrared camera S1 and the still camera S2 at step S301 without performing the face recognition.
As described above, the present invention captures face images by simultaneously using an infrared camera and a still camera, determines whether a current face is a forged face by analyzing the characteristics of an input infrared image, and performs user authentication using a photorealistic image captured by visible light if it is determined that the current face is not a forged face, thus improving the reliability and security of a face detection device.
While the invention has been shown and described with respect to the preferred embodiments, it will be understood by those skilled in the art that various changes and modifications may be made without departing from the scope of the invention as defined in the following claims.
Number | Date | Country | Kind |
---|---|---|---|
10-2008-0117354 | Nov 2008 | KR | national |
Number | Name | Date | Kind |
---|---|---|---|
20020136435 | Prokoski | Sep 2002 | A1 |
20080144893 | Guo et al. | Jun 2008 | A1 |
Number | Date | Country |
---|---|---|
2008-181468 | Aug 2008 | JP |
10-2008-0037447 | Apr 2008 | KR |
Entry |
---|
Jiangwei Li et al., “Live Face Detection Based on the Analysis of Fourier Spectra”, National Laboratory of Pattern Recognition, Institute of Automation, Chinese Academy of Science, Beijing. |
Number | Date | Country | |
---|---|---|---|
20100128938 A1 | May 2010 | US |