KR100842258B1 - Methdo for detecting forged face image and apparatus thereof - Google Patents

Abstract

According to an embodiment of the present invention, a method and apparatus for determining whether a face image is forged include detecting a center of two eyes in a continuously input face image; Normalizing the size and illumination of the face image based on the centers of the two eyes; And detecting whether the face image is forged by detecting eye movement after detecting two eye regions in the normalized face image, and additionally included in the existing resource-constrained embedded face recognition system. It can improve the security and reliability of the face recognition system by preventing fraudulent use by discriminating fake and real faces such as pictures with a small amount of calculation without equipment.

Face recognition, fake face discrimination, binarization, hamming distance

Description

Method and apparatus for detecting forgery of face image {Methdo for detecting forged face image and apparatus}

1 is a flowchart illustrating an entire face recognition process to which a method for determining whether a face image is forged according to the present invention is applied.

FIG. 2 is a flowchart illustrating the fake face determination process 120 of FIG. 1 in more detail.

3 is a diagram illustrating facial region normalization in a method of determining whether a face image is forged according to the present invention.

4 is a diagram illustrating an eye region binarization image in a method of determining whether a face image is forged according to the present invention.

FIG. 5 is a flowchart illustrating a forgery face discrimination process 260 using a hamming distance in detail in the method of determining whether a face image is forged according to the present invention.

6 is a block diagram illustrating a configuration of an apparatus for determining whether a face image is forged.

The present invention relates to a method and apparatus for determining whether an input face image is a real face or a photographic image in a user authentication system using a face. More particularly, the present invention relates to detecting eye regions from continuously input face images. The present invention relates to a method and apparatus for determining whether or not a fake image is forged by calculating a change amount of detected eye regions and determining whether it is a real face or a photographic image.

Bio-recognition technology has the advantage of not needing to be lost or memorized because it is a personal authentication technology using information on individual body characteristics and behaviors. It has the advantage of being safer than the technology used, so it is being spotlighted as the next generation technology to replace the existing password and ID card-based personal authentication technology.

However, as forgery bio-information generation technology in bio-recognition also develops, the problem of counterfeit bio-information generated when inputting bio-information into the system becomes the biggest factor that degrades the reliability and security of bio-recognition. Algorithms have difficulty distinguishing real bio information from counterfeit bio information.

In the case of face recognition, the fake face and the real face can be easily distinguished by using the vein map of the face taken by the ultraviolet camera or by using the heat distribution of the face taken by the thermal infrared camera, but these camera equipment is very expensive. There is a limit to the practical application. In addition, the method of instructing the user to speak or move has a problem of making the user feel uncomfortable.

The technical problem to be achieved by the present invention has been proposed to solve the above-mentioned conventional problems. In the resource-constrained embedded facial recognition system, a memory usage and a calculation complexity can be reduced using only a conventional camera without using additional equipment. In addition, the present invention provides a method and apparatus for determining whether or not a forgery of a face image that can increase the reliability and security of the face recognition system by determining whether the input image is a fake face or a real face, such as a face photograph.

In order to achieve the above technical problem, a method for determining whether a face image is forged according to the present invention comprises the steps of detecting the center of the two eyes in the face image is continuously input; Normalizing the size and illumination of the face image based on the centers of the two eyes; And determining whether the face image is forged by detecting eye movements after detecting two eye regions in the normalized face image.

In order to achieve the above technical problem, the apparatus for determining whether or not the face image according to the present invention comprises an image acquisition unit for continuously obtaining a face image; A normalizer for performing normalization on the size and lighting of the acquired face image based on the centers of two eyes; An extraction unit for extracting and binarizing an eye region from a normalized face image; And a determining unit configured to calculate whether the binarized eye region is changed and compare the value with a predetermined threshold to determine whether the face image inputted is a forgery.

Hereinafter, a fake face discrimination method according to the present invention will be described in detail with reference to the accompanying drawings. FIG. 1 is a flowchart illustrating an entire face recognition process to which a method for determining whether a face image is forged according to the present invention is applied, and FIG. 2 is a flowchart illustrating the fake face determination process 120 of FIG. 1 in more detail. 3 is a diagram illustrating facial region normalization in a method of determining whether a face image is forged according to the present invention, and FIG. 4 is a diagram illustrating eye region binarization in a method of determining whether a face image is forged according to the present invention. A diagram showing an image. 5 is a flowchart illustrating a forgery face determination process 260 using a hamming distance in detail in the method of determining whether a face image is forged according to the present invention. 6 is a block diagram illustrating a configuration of an apparatus for determining whether a face image is forged.

First, the overall flow will be described with reference to FIG. 1. As shown in FIG. 1, the entire face recognition process receives a continuous image (110) and determines whether it is a fake face (120). In the fake face discrimination process 120, it is determined whether the input image is a fake face or a real face (130). If the actual face is subjected to a full face recognition process (140), in the case of a fake face face recognition is terminated.

Now look at the fake face discrimination process according to the present invention with reference to FIG. There is almost no movement of facial expressions or facial components in the sequential face images, but the area with the most movement is the eye area. The eye area consciously or unconsciously causes blinking or eye movement. Therefore, in the present invention, a method of detecting a fake face by detecting eye movement was used.

When the image is acquired (210), the initial face region is detected from the image, and the centers of the two eyes are detected in the detected face region (220).

Next, the face region is normalized 230, and size normalization and illumination normalization are performed. First, size normalization is performed on the face area by using the centers of the two eyes detected in step 220. Since the size or direction of the object to be recognized may be different, it is intended to facilitate recognition by normalizing it to the same size and direction. Next, the normalization of the size normalized face region is performed by using a method such as histogram smoothing to reduce the influence on the illumination. The reason for the preprocessing of the illumination is to prevent the pixel value from being changed by the illumination in the process of binarizing the eye area later.

Then, a predetermined eye region is extracted based on the centers of the two eyes from the face image normalized with respect to the size and illumination, and a predetermined threshold value (where the threshold value can be arbitrarily set by the user according to the situation and can be set by trial and error method) After the binarization is performed using the storage 240, the process is repeated N times (250). N times, N left eye regions (LeftEye1, 2, ..., N) and N right eye regions (RightEye1, 2, ..., N) are generated. At this time, the number of repetition can be arbitrarily set by the user according to the situation, and it is desirable to obtain the most appropriate number of times by trial and error method.

Finally, after comparing the stored eye areas, a change amount is calculated to determine whether or not a fake face is present (260).

Therefore, it is possible to determine whether a real face or a fake face using a photo in real time in the embedded environment through a small amount of calculation without additional equipment or user inconvenience.

3 illustrates a process 230 of normalizing a face region with respect to size and illumination in the above-described fake face determination process, and normalization of the face region may be performed as follows.

Looking at the two eyes of the detected face image, the head is inclined as shown in 310, and thus the line connecting the two eyes is inclined. By using the detected tilt of the center of the two eyes, the image is rotated to correct the tilt of the image, and the distance between the centers of the two eyes by a certain size (where the size can be arbitrarily set by the user according to the situation. A face image is created by clipping the face candidate area of the face candidate region. In operation 320, the clipped face image is sampled at a predetermined reference size (in which the reference size may be arbitrarily set by the user and may be set in a trial and error method).

Finally, in order to reduce the change in illumination of the sampled face region, the lighting normalization process is performed in the same manner as in the histogram smoothing (330).

FIG. 4 is a diagram illustrating a result of binarizing an eye region in the above-described forgery face discrimination process, and binarization of the eye region may be performed as follows.

The eye region is extracted based on the centers of the two eyes from the face image normalized in step 230. Each pixel value of the gray image is converted into a binary image of 0 and 1 using a predetermined threshold value (the threshold value can be arbitrarily set by the user and can be set by trial and error method). As shown in the figure, the center of the eye is changed not only by the blink of the eye but also by the movement of the eye. As shown in the figure, since the center of the eye is always detected, the binary image of the continuous face is almost similar (410). In the case of a real face, a difference occurs between binary images of an eye region due to movement of a pupil or blink of an eye (420).

FIG. 5 is a detailed flowchart illustrating a process 260 of comparing the binarized eye images to calculate a change amount and determining whether a fake face is present in the fake face discrimination method according to the present invention. First, the binarized N left eyes and N right eyes are input (510) and compared using a Hamming distance method (520). The total number of comparisons adds both left and right eyes to N (N-1). It is time. Hamming distance is a similarity comparison method that indicates the number of different digits by comparing each corresponding digit of two binary numbers with the same number of bits. After adding all the changes calculated using the Hamming distance, the change amount is compared with the predetermined threshold value (where the threshold value can be set by the user arbitrarily and can be set by trial and error method) (530). If it is small, it is determined as a real face (550). If it is larger than the threshold, it is determined as a fake face (540).

Now, an embodiment of an apparatus for determining whether a face image is forged according to the present invention will be described with reference to FIG. 6. First, the image acquisition unit 610 continuously receives a face image provided by an external recognition device. The normalizer 620 performs a function of normalizing size and lighting based on the centers of two eyes with respect to the face image. The normalization unit 620 will be described in detail. The inclination of the line connecting two eyes is obtained from the face image to correct the inclination of the image to make the face image horizontal. The clipping unit 623 clips a face area of a predetermined size with respect to the face image whose slope is corrected, and then samples and outputs a predetermined size.

The extractor 630 extracts, binarizes, and outputs an eye region from the normalized face image. The determination unit 640 calculates a change amount of the binarized eye region output by the extraction unit 630 and compares the value with a predetermined threshold to determine whether the input face image is forged. At this time, the determination unit compares each of the binarized N (N is a natural number) left eyes and right eyes, and then determines whether the face image is forged by comparing a change amount of the eye region with a predetermined threshold value. The thresholds used in the description of FIG. 6 and the functions at each component are the same as those described above with reference to FIGS.

The method for determining whether a face image is forged according to the present invention can also be embodied as computer readable codes on a computer readable recording medium. The computer-readable recording medium includes all kinds of recording devices in which data that can be read by a computer system is stored. Examples of computer-readable recording media include ROM, RAM, CD_ROM, magnetic tape, floppy disks, and optical data storage, and may also include those implemented in the form of carrier waves (e.g., transmission over the Internet). . The computer readable recording medium can also be distributed over network coupled computer systems so that the computer readable code is stored and executed in a distributed fashion.

As described above, optimal embodiments have been disclosed in the drawings and the specification. Although specific terms have been used herein, they are used only for the purpose of describing the present invention and are not intended to limit the scope of the present invention as defined in the claims or the claims. Therefore, those skilled in the art will understand that various modifications and equivalent other embodiments are possible from this. Therefore, the true technical protection scope of the present invention will be defined by the technical spirit of the appended claims.

As described above, the method and apparatus for determining a fake image of a face image according to the present invention are negated by discriminating a fake face such as a photo and a real face with a small amount of computation without additional equipment in the existing resource-constrained embedded face recognition system. By preventing its use, it can improve the security and reliability of facial recognition system.

Claims (9)

(a) detecting the centers of the two eyes in the continuously input face images; (b) normalizing the size and illumination of the face image based on the centers of the two eyes; And and (c) determining whether the face image is forged by detecting eye movement after detecting two eye regions in the normalized face image. The method of claim 1, wherein step (b) (b1) correcting the inclination by rotating the face image according to the inclination of the center of the line connecting the two eyes; (b2) clipping the face candidate area using the center distance of the two eyes; (b3) normalizing by sampling the clipped face candidate area to a predetermined size; And (b4) performing normalization on illumination with respect to the normalized face region; and determining whether the face image is forged. The method of claim 1, wherein step (c) (c1) extracting and binarizing a constant eye region based on the centers of the two eyes; And (c2) determining whether the face image is a real face image by calculating a change amount of the binarized eye region. The method of claim 3, wherein step (c1) extracting a predetermined eye region from the normalized face image based on the centers of the two eyes; And and (c2) binarizing the extracted eye region by applying a predetermined threshold value. The method of claim 3, wherein step (c2) (c21) comparing each of the binarized N (N is natural numbers) left eyes and right eyes, and then determining whether the face image is forged by comparing a change amount of the eye region with a predetermined threshold; Method for determining whether the forgery of the facial image, characterized in that it comprises a. The method of claim 5, And comparing each of the binarized N left and right eyes using a Hamming distance calculation method. An image acquisition unit for continuously acquiring a face image; A normalizer for performing normalization on the size and lighting of the acquired face image based on the centers of two eyes; An extraction unit for extracting and binarizing an eye region from a normalized face image; And A determination unit for determining whether the input face image is forged by calculating a change amount of the binarized eye region and comparing the value with a predetermined threshold value to determine whether the face image is forged Device. The method of claim 7, wherein the normalization unit A tilt correction unit for correcting a tilt of an image based on a slope of a line segment connecting the two eyes; And And a clipping unit for clipping a face area of a predetermined size with respect to the face image whose slope is corrected and sampling and outputting the sample to a predetermined size. The method of claim 7, wherein the determining unit Comparing the left and right eyes of each binarized N (N is a natural number), and comparing the amount of change in the eye region with a predetermined threshold to determine whether the face image is forged or not. Device for determining forgery.

Images