KR100707195B1 - Method and apparatus for detecting classifier having texture information of face, Method and apparatus for recognizing face using statistical character of the texture information - Google Patents

Abstract

Disclosed are a classifier detection method and apparatus having facial texture information and a face recognition method and apparatus using statistical characteristics of texture information. The classifier detection method having texture information of a face includes detecting a first image and a second image of the same face, dividing the detected first image and the second image into a predetermined size, respectively, and partial images of the divided first image. Detecting first partial images corresponding to the first partial images and second partial images corresponding to the divided second images, and first textures corresponding to texture information of each of the detected first partial images. Generating second texture information corresponding to texture information of each of the pieces of information and the detected second partial images, and comparing the similarity of each of the first texture information with the second texture information corresponding to the first texture information. Detecting, from the first partial images, first classifiers capable of recognizing the identity of the face, in accordance with the inspecting and the similarity inspected. It shall be. Therefore, according to the present invention, in face recognition, since face identity is determined using texture information of a face, face recognition can be accurately performed.

Description

Method and apparatus for detecting classifier having texture information of face, method and apparatus for recognizing face using statistical character of the texture information}

1 is a flowchart of an embodiment for explaining a classifier detection method having texture information of a face according to the present invention.

FIG. 2 is a flowchart of an exemplary embodiment for explaining the fourteenth step shown in FIG. 1.

FIG. 3 is a diagram for describing, for example, a method of detecting the first classifiers shown in FIG. 1.

FIG. 4 is a diagram for explaining, for example, a method of detecting the second classifiers shown in FIG. 1.

5 is a flowchart of an embodiment for describing a face recognition method using statistical characteristics of texture information according to the present invention.

FIG. 6 is a flowchart of an exemplary embodiment for describing operation 54 of FIG. 5.

7 is a block diagram of an exemplary embodiment for explaining a classifier detection apparatus having face texture information according to the present invention.

FIG. 8 is a block diagram of an exemplary embodiment for describing the texture information generator illustrated in FIG. 7.

9 is a block diagram of an embodiment for describing a face recognition apparatus using statistical characteristics of texture information according to the present invention.

FIG. 10 is a block diagram of an exemplary embodiment for describing the texture information generator illustrated in FIG. 9.

<Brief description of the major symbols in the drawings>

100: face image detector 110: partial image detector

120: texture information generation unit 130: similarity inspection unit

140: first classifier detector 150: second classifier detector

200: information detector 210: histogram

300: face image detector 310: partial image detector

320: texture information generation unit 330: similarity inspection unit

340: Face recognition unit 400: Information detector

410: histogram

The present invention relates to a face recognition technology among biometrics technology, and more particularly, to a classifier detection method and apparatus for face recognition using a statistical property of texture information on a face, and a face using a classifier. A recognition method and apparatus are provided.

In recent years, it is necessary to use an access card to identify the visitors of the building, but the conventional method has been to contact a card reader with a key card or key pad assigned to each individual. I am using my method of identification. This method can cause security problems in that it is inconvenient to carry a key card or keypad with you at all times, and that another person can be allowed to enter in case of theft or loss.

In recent years, in order to minimize such side effects and dysfunction, the biometric technology that automatically recognizes or verifies an individual's identity using human's own biological or behavioral characteristics of human beings has been in the spotlight. A reliable biometric system, in addition to the password replacement effect of ATM cards and the access control system of buildings, requires high levels of attention, such as bank vaults, between companies and airport security systems, between future humans and machines. It is a trend to have a general range of applications, such as the interface of the, and many studies have been made to make it more applicable and have higher reliability.

Biometric System refers to a personal identification and authentication system using physical characteristics. The Association of Biometric Technologies defines biometric technology as 'a study of measurable physical or personal characteristics in order to verify or identify an individual's characteristics by automated means.' The personal characteristics of these biometrics cannot be communicated by theft or leakage, and there is no risk of change or loss. Therefore, the use of these techniques allows the audit function to be fully tracked. The advantage is that it can be built.

Personal characteristics used in implementing such a biometric system include fingerprint, face, palmprint, hand geometry, thermal image, voice, and handwriting. Signature, Vein, Typing keystroke dynamics, Retina and Iris.

In particular, the face recognition technology is the most natural and relatively low biometric recognition technology because the shape of the face that humans most often use when identifying another person is the shape of the face.

However, in the conventional face recognition technology, by comparing features such as contours of the face of the face to determine whether they are identical, people who are not the same person are recognized as the same person, or the same person is affected by the influence of lighting, facial expression, posture, etc. There is a problem that frequently makes an error in face recognition, such as recognizing that they are not the same.

An object of the present invention is to provide a method for detecting a classifier having facial texture information.

Another object of the present invention is to provide a method for recognizing the identity of a face by checking similarity of texture information of a detected classifier.

Another object of the present invention is to provide an apparatus for detecting a classifier having texture information of a face.

Another object of the present invention is to provide an apparatus for recognizing whether or not a face is identical by checking similarity of texture information of a detected classifier.

In order to achieve the above object, the classifier detection method having the texture information of the face according to the present invention comprises the steps of detecting the first image and the second image of the same face, each of the detected first image and the second image of a predetermined size Dividing and detecting first partial images corresponding to partial images of the divided first image and second partial images corresponding to partial images of the divided second image, respectively; detected first partial images Generating second texture information corresponding to the texture information of each of the first texture information and the detected second partial images corresponding to the respective texture informations, in the first texture information and the first texture information. A first partial image of the first classifiers capable of recognizing the identity of the face, according to the similarity of each of the corresponding second texture informations and the examined similarity. In that it comprises a step of detecting from the features.

In order to achieve the above object, the face recognition method using the statistical characteristics of the texture information according to the present invention comprises the steps of detecting the image of the face, classifying the detected image to a predetermined size, the recognition of the identity of the face among the divided parts Detecting partial images corresponding to classifiers useful for the object, generating texture information of each of the detected partial images, and checking similarity between generated texture information and texture information of a previously stored face. According to the similarity, characterized in that it comprises the step of recognizing the identity of the face.

In order to achieve the above object, the classifier detecting apparatus having the texture information of the face according to the present invention is a face image detecting unit for detecting the first image or the second image of the same face, the detected first image or the second image. A partial image detector which detects the first partial images corresponding to the partial images of the divided first image or the second partial images corresponding to the partial images of the divided second image, divided into a predetermined size, and the detected first image A texture information generator for generating first texture information corresponding to the texture information of each of the partial images or second texture information corresponding to the texture information of each of the detected second partial images, the first texture information, and According to the similarity checker for checking the similarity of each of the second texture information corresponding to the first texture information and the checked similarity, And a first classifier detector configured to detect first classifiers capable of recognizing the same from the first partial images.

In order to achieve the above object, the face recognition apparatus using the statistical characteristics of the texture information according to the present invention comprises a face image detection unit for detecting the image of the face, the detected image is divided into a predetermined size, and the A partial image detector for detecting partial images corresponding to classifiers useful for identity recognition, a texture information generator for generating texture information of each of the detected partial images, and the similarity between the generated texture information and texture information of a previously stored face. And a face recognition unit that recognizes whether or not the face is identical according to the similarity inspection unit for inspecting and the similarity inspected.

Hereinafter, a classifier detection method having a face texture information according to the present invention will be described in detail with reference to the accompanying drawings.

1 is a flowchart of an embodiment for explaining a classifier detection method having texture information of a face according to the present invention.

First, a first image and a second image of the same face are detected (step 10). When detecting two images, ie, first and second images, respectively for the same face, it is preferable to detect in the same direction of the face.

When detecting images of a face other than the front face, generalizing the face images toward the front face with respect to the face images other than the front face based on the position of the eye of the face.

The first image and the second image may be filtered using a Gaussian low pass filter. By filtering the first image and the second image, fine noise or the like appearing on the face may be removed.

After the tenth step, the detected first image and the second image are respectively divided into predetermined sizes, and the first partial images corresponding to the partial images of the divided first image and the partial images of the divided second image respectively. Detecting second partial images, respectively (Step 12). A window having a predetermined size is used to detect a plurality of first partial images from the first image. In addition, a window having a predetermined size is used to detect a plurality of second partial images from the second image. If the size of the first and second images is 130 [pixel] x 150 [pixel], a plurality of first and second partial images having a size of 20 [pixel] x 20 [pixel] which is a predetermined window size Detect each one.

Meanwhile, the first partial images to be detected are characterized in that predetermined portions of the image overlap each other. For example, the images are overlapped with each other by a predetermined pixel. Thus, partial images adjacent to each other share the same image in the overlapping area.

The second partial images may also be overlapped with predetermined portions of the image.

After the twelfth step, the first texture information corresponding to the texture information of each of the detected first partial images and the second texture information corresponding to the texture information of each of the detected second partial images are generated. Step 14).

FIG. 2 is a flowchart of an exemplary embodiment for explaining the fourteenth step shown in FIG. 1.

First, first texture information and second texture information are detected from the first partial images and the second partial images, respectively, using a local binary pattern (LBP) method (Step 30).

In particular, Haar morphology wavelet method, median morphology wavelet method, Erodent morphology wavelet method and expanded morphology wavelet among local binary pattern methods The first texture information and the second texture information are detected using any one of the methods.

The morphology wavelet method refers to a method of detecting desired information using a morphology operation in a specific digital signal. This morphology wavelet method is a general concept and will not be described in detail below, and briefly describes the detection of texture information by the lower morphology wavelet method. The lower morphology wavelet method uses Equation 1 below.

S _n = min [x _2n , x _{2n + 1} ]

d _n = x _2n- x _{2n + 1}

Here, x _2n and x _{2n + 1} represent pixel pixel values, respectively, S _n means a minimum pixel value of x _2n and x _{2n + 1} , and d _n represents x _2n and It means the difference of the pixel value of x _{2n + 1} .

Texture information is detected by repeatedly calculating the calculation in the horizontal and vertical directions of the partial image using the above-described equation.

After the thirtieth step, the detected first texture information and the second texture information are histogram respectively (step 32). The number of pixels is histogram according to the brightness for each of the detected first texture information and the second texture information.

The horizontal axis represents brightnesses divided into a predetermined size (eg, brightness divided into 256 levels), and the vertical axis represents the number of pixels of each brightness included in one texture information.

After the fourteenth step, the similarity between each of the first texture information and the second texture information corresponding to the first texture information is examined (step 16). In step 30, the histogramized first texture information and the second texture information are compared with each other to check their similarities. That is, the similarity is checked by comparing the number of pixels according to the brightness of the specific texture information among the first texture information and the number of pixels according to the brightness of the texture information corresponding to the specific texture information among the second texture information.

In this manner, all similarities between the first texture information and the second texture information corresponding to the first texture information are examined.

Specifically, similarity is checked using any one of the following methods: chi square distance, Kullback-Leibler distance, and Jensen-Shannon distance. It is characterized by. The histogram is used to determine the similarity of texture changes in an image. To compare the similarity of these histograms, one of the following measures is used: Chi square distance, Kullback-Leibler distance, and Jensen-Shannon distance.

To determine the similarity using the chi square length, Equation 2 below is used.

Here, S _i represents the number of pixels of the i-th brightness of the specific texture information among the first texture information, and M _i represents the number of pixels of the i-th brightness of the texture information corresponding to the above-mentioned specific texture information among the second texture information. Indicates.

Meanwhile, the following Equation 3 or 4 is used to determine the similarity using the coolback-labeler length.

Here, S _i represents the number of pixels of the i-th brightness of the specific texture information among the first texture information, and M _i represents the number of pixels of the i-th brightness of the texture information corresponding to the above-mentioned specific texture information among the second texture information. Indicates.

Meanwhile, the following Equation 5 is used to determine the similarity using the Jensen-Shonon length.

Here, S _i represents the number of pixels of the i-th brightness of the specific texture information among the first texture information, and M _i represents the number of pixels of the i-th brightness of the texture information corresponding to the above-mentioned specific texture information among the second texture information. Indicates.

The smaller the size of the chi square length, coolback-labeller length, or Jensen-Shonon length obtained from the histogramized texture information, indicates that the first image and the second image are similar.

The chi square length, the coolback-liver length, or the Jensen-Sennon length for all of the first texture information and all of the corresponding second texture information are obtained. These values for each texture information are used for detection of the first classifiers, which will be described later.

After the sixteenth step, according to the examined similarity, first classifiers capable of recognizing the identity of the face are detected from the first partial images (step 18).

Examine similarity for each texture information using any one of the aforementioned Chi square distance, Kullback-Leibler distance, and Jensen-Shannon distance. As a result, texture information whose checked value is equal to or less than a predetermined threshold is detected as first classifiers.

As a result of examining the similarity of each of the first texture information and the second texture information, partial images of the texture information whose magnitude of the detected chi-square length, coolback-labeller length, or jensen-shannon length are below a predetermined threshold are classified into a first classification. Detects as those. That is, the first classifier extracts partial images that are effective to recognize the sameness of the face among the first partial images.

FIG. 3 is a diagram for describing, for example, a method of detecting the first classifiers shown in FIG. 1. First, a first image and a second image of the same face are detected. The first partial images and the second partial images corresponding to the partial images of each of the detected first and second images are detected. First texture information of each of the detected first partial images is detected, and second texture information of each of the second partial images is detected. Histograms each of the detected first texture information and histograms each of the second texture information. The similarity between the histogramized first texture information and the second texture information is examined, and the first classifiers effective for recognition of the identity of the face are detected from the examined similarity.

Meanwhile, another first classifier is detected while repeating the aforementioned steps 12 to 18 while varying the window size with respect to the partial images detected from the first image and the second image. This makes it possible to detect first classifiers of varying window sizes.

After the sixteenth step, a second classifier among the detected first classifiers, which is useful for recognizing the identity of the face, is detected using a Bayesian network method (step 20). Bayesian networks are widely used to infer the purpose of software users as a tool for probabilistic modeling of cause-effect relationships between variables.

The generated first classifiers are grouped by classifiers with the same window size, the Bayesian network is used to learn the respective confidence values of the bundled first classifiers, and the learned confidence values are first classified. The second classifiers are detected by multiplying by.

FIG. 4 is a diagram for explaining, for example, a method of detecting the second classifiers shown in FIG. 1. As shown in FIG. 4, among the first classifiers having the same window size among various first classifiers having different window sizes detected by repetition of the twelfth to sixteenth steps, the first classifiers are divided and classified. Among the classifiers, each second classifier is detected using a Bayesian network. The detected second classifiers are used in the face recognition method described later.

Hereinafter, a face recognition method using statistical characteristics of texture information according to the present invention will be described in detail with reference to the accompanying drawings.

5 is a flowchart of an embodiment for describing a face recognition method using statistical characteristics of texture information according to the present invention.

First, an image of a face is detected (50th step). When detecting an image of a face, when detecting an image of a face other than the front face, it is generalized to a face image facing the front with respect to the face image not the front face based on the position of the eye of the face.

Filtering the detected image using a Gaussian low pass filter. By filtering the image, fine blemishes and the like appearing on the face can be removed.

After operation 50, partial images corresponding to classifiers useful for recognizing identity of a face are detected from the detected image (operation 52).

Information about classifiers useful for face recognition is provided as information after being detected in advance from the classifier detection method for face recognition shown in FIG. In particular, the second classifiers detected by the Bayesian Network method as classifiers are used as classifiers useful for face recognition.

The detected partial images are characterized in that predetermined portions of the image overlap each other. For example, the images are overlapped with each other by a predetermined pixel. Thus, partial images adjacent to each other share the same image in the overlapping area.

After operation 52, texture information of each of the detected partial images is generated (operation 54).

FIG. 6 is a flowchart of an exemplary embodiment for describing operation 54 of FIG. 5.

First, texture information is detected from each of the divided images by using a local binary pattern (LBP) method (step 70).

In particular, Haar morphology wavelet method, median morphology wavelet method, Erodent morphology wavelet method and expanded morphology wavelet among local binary pattern methods The first texture information and the second texture information are detected using any one of the methods.

After operation 70, the detected texture information is histogramd (operation 72).

The number of pixels is histogramized according to the brightness of the detected texture information. The horizontal axis represents brightnesses divided into a predetermined size (eg, brightness divided into 256 levels), and the vertical axis represents the number of pixels of each brightness included in one texture information.

After operation 54, a similarity between the generated texture information and texture information of a previously stored face is inspected (operation 56). The histogramized texture information generated in step 70 is compared with the similarity between the histogramized texture information stored in a predetermined storage space in advance for face identification.

Specifically, similarity is checked using any one of the following methods: chi square distance, Kullback-Leibler distance, and Jensen-Shannon distance. It is characterized by.

After the fifty sixth step, according to the checked similarity, whether the face is the same or not (step 58).

The similarity for each texture information is determined using any one of the aforementioned Chi square distance, Kullback-Leibler distance, and Jensen-Shannon distance. As a result of the inspection, when the average of the measured values is equal to or less than a predetermined threshold for determining the identity of the face, the face in which the image is detected is recognized as the same person as the previously stored face. However, if the average of the examined values is not less than a predetermined threshold, it is recognized that the person is not the same person as the previously stored face. However, the method of recognizing the identity of the face by comparing the average of the inspected values with a predetermined threshold is just one example, and there may be other modifications.

Hereinafter, a classifier detection apparatus having a face texture information according to the present invention will be described in detail with reference to the accompanying drawings.

FIG. 7 is a block diagram of an exemplary classifier detecting apparatus having face texture information according to an embodiment of the present invention. The face image detector 100, the partial image detector 110, the texture information generator 120, The similarity checker 130, the first classifier detector 140, and the second classifier detector 150 are configured.

The face image detector 100 detects the first image or the second image of the same face, and outputs the detected result to the partial image detector 110.

The face image detector 100 detects the first image or the second image in a direction in which the face faces the front.

The face image detector 100 may filter the first image or the second image using a Gaussian low pass filter. The face image detector 100 filters the first image and the second image to remove fine blemishes and the like that appear on the face.

The partial image detector 110 divides the detected first image or the second image into a predetermined size, and divides the detected first image or the second image into first partial images corresponding to the partial images of the divided first image or partial images of the divided second image. The corresponding second partial images are detected. The partial image detector 110 uses a window having a predetermined size to detect the plurality of first partial images from the first image. In addition, the partial image detector 110 uses a window having a predetermined size to detect a plurality of second partial images from the second image.

The partial image detector 110 detects the images such that the first partial images overlap each other by a predetermined portion, or detects the images so that the second partial images overlap each other by a predetermined portion. When the partial image detector 110 detects that the images overlap each other by a predetermined pixel, the partial images adjacent to each other share the same image in the overlapping area.

The texture information generator 120 generates first texture information corresponding to texture information of each of the first partial images detected by the partial image detector 110, or the second portion detected by the partial image detector 110. Second texture information corresponding to texture information of each image is generated.

FIG. 8 is a block diagram illustrating an example of the texture information generator 120 illustrated in FIG. 7, and includes an information detector 200 and a histogram 210.

The information detector 200 detects first texture information from first partial images or second texture information from second partial images by using a local binary pattern (LBP) method.

The information detecting unit 200 may include a lower morphology wavelet method, a median morphology wavelet method, an erodent morphology wavelet method, and an extended morphology wavelet method among local binary pattern methods. expanded morphology wavelet).

The histogram unit 210 histograms the first texture information or the second texture information detected by the information detector 200. The histogram unit 210 histograms the number of pixels according to brightness with respect to each of the detected first texture information and the second texture information. The histogramized texture information indicates brightnesses in which the horizontal axis is divided into predetermined sizes (eg, brightness divided in 256 steps), and the vertical axis indicates the number of pixels of each brightness included in one texture information.

The similarity checking unit 130 compares the first texture information and the second texture information histogramized by the histogram unit 210 and checks each similarity. That is, the similarity inspection unit 130 compares the number of pixels according to the brightness of the specific texture information among the first texture information and the number of pixels according to the brightness of the texture information corresponding to the specific texture information among the second texture information to compare the similarity. Check it. The similarity checking unit 130 examines both similarities of the first texture information and the second texture information corresponding to the first texture information in this manner.

In particular, the similarity inspection unit 130 measures any one of Chi square distance, Coolback-Leibler distance, and Jensen-Shannon distance by checking similarity. To examine similarity. The similarity checking method using the Chi square distance, the Kullback-Leibler distance, and the Jensen-Shannon distance is the same as described above, and thus description thereof is omitted.

The first classifier detector 140 detects first classifiers capable of recognizing the sameness of the face from the first partial images, according to the similarity inspected by the similarity checker 130. Examine similarity for each texture information using any one of the aforementioned Chi square distance, Kullback-Leibler distance, and Jensen-Shannon distance. Upon receiving a result, the first classifier detector 140 detects, as the first classifier, texture information whose checked value is equal to or less than a predetermined threshold.

The second classifier detector 150 detects the second classifiers useful for the recognition of the same face among the detected first classifiers using a Bayesian network method.

Hereinafter, a face recognition apparatus using statistical characteristics of texture information according to the present invention will be described in detail with reference to the accompanying drawings.

FIG. 9 is a block diagram illustrating an example of a face recognition apparatus using statistical characteristics of texture information according to the present invention. The face image detector 300, the partial image detector 310, the texture information generator 320, It is composed of a similarity inspection unit 330 and the face recognition unit 340.

The face image detector 300 detects an image of a face and outputs the detected result to the partial image detector 310.

The face image detector 300 may detect an image of a direction in which the face faces the front.

The face image detector 300 may filter the image using a Gaussian low pass filter. The face image detector 300 filters the image to remove fine blemishes and the like that appear on the face.

The partial image detector 310 detects partial images corresponding to classifiers useful for recognizing the identity of a face from the detected image, and outputs the detected result to the texture information generator 320. The partial image detector 310 includes information on classifiers useful for recognizing the identity of a face previously detected by the classifier detection apparatus for face recognition of FIG. 7. In particular, the partial image detector 310 uses the second classifiers detected by the Bayesian Network method as classifiers as classifiers useful for face recognition.

The partial image detector 310 detects the images such that the partial images overlap each other by a predetermined portion.

The texture information generator 320 generates texture information of each of the partial images detected by the partial image detector 310, and outputs the generated result to the similarity checker 330.

FIG. 10 is a block diagram illustrating an example of the texture information generator 320 illustrated in FIG. 9, and includes an information detector 400 and a histgram 410.

The information detector 400 detects texture information from the partial images by using a local binary pattern (LBP) method.

In particular, the information detector 400 may include a lower morphology wavelet method, a median morphology wavelet method, an etchant morphology wavelet method, and an extended morphology wave among local binary pattern methods. It characterized by using any one of the (expanded morphology wavelet) method.

The histogram unit 410 histograms the detected texture information. The histogram unit 410 histograms the number of pixels according to brightness with respect to each of the detected first texture information and the second texture information. The histogramized texture information indicates brightnesses in which the horizontal axis is divided into predetermined sizes (eg, brightness divided in 256 steps), and the vertical axis indicates the number of pixels of each brightness included in one texture information.

The similarity inspection unit 330 checks the similarity between the generated texture information and the texture information of the previously stored face. The similarity checker 330 compares the histogramized texture information generated by the histgramming unit 410 with the similarity of the histogramized texture information previously stored in a predetermined storage space to recognize the identity of the face.

The similarity inspection unit 330 checks the similarity by using any one of Chi square distance, Kullback-Leibler distance, and Jensen-Shannon distance. It is characterized by.

The face recognition unit 340 recognizes whether a face is identical according to the similarity inspected by the similarity inspection unit 330.

Examine similarity for each texture information using any one of the aforementioned Chi square distance, Kullback-Leibler distance, and Jensen-Shannon distance. As a result, when the average of the measured values is equal to or less than a predetermined threshold for determining the identity of the face, the face recognizing unit 340 recognizes the face in which the image is detected as the same person as the previously stored face. However, if the average of the inspected values is not less than a predetermined threshold, the face recognizing unit 340 recognizes that the face is not the same person as the previously stored face.

The face recognition unit 340 recognizes the sameness of the face by comparing the average of the measured values with a predetermined threshold, and is only an example. Can be.

Meanwhile, the above-described embodiments of the present invention may be embodied as computer readable codes / instructions / programs, and the codes / instructions may be implemented using a medium, for example, a computer readable recording medium. / Can be implemented in a general-purpose digital computer for operating the program. The computer-readable recording medium may be a magnetic storage medium (eg, ROM, floppy disk, hard disk, magnetic tape, etc.), optical reading medium (eg, CD-ROM, DVD, etc.) and carrier wave (eg Storage media). In addition, embodiments of the present invention may be implemented as a medium (s) containing computer readable code, such that a plurality of computer systems connected through a network may be distributed and processed. Functional programs, codes and code segments for realizing the present invention can be easily inferred by programmers in the art to which the present invention belongs.

Such a classifier detection method and apparatus having texture information of the present inventors and a face recognition method and apparatus using statistical characteristics of texture information have been described with reference to the embodiments shown in the drawings for clarity, but this is merely illustrative. Those skilled in the art will appreciate that various modifications and equivalent other embodiments are possible therefrom. Therefore, the true technical protection scope of the present invention will be defined by the appended claims.

In the face recognition method and apparatus using the statistical characteristics of the texture information according to the present invention, in face recognition, by determining the identity of the face using the texture information of the face, it recognizes a person who is not the same person, or the same person This function prevents errors in face recognition, such as recognizing that they are not the same by the influence of lighting, facial expression, posture, and the like.

The classifier detection method and apparatus having the texture information of the face according to the present invention can effectively classify classifiers useful for face recognition and use these classifiers for face recognition to determine whether the face is identical. This has the effect of allowing a quick conclusion to be reached.

Claims (33)

(a) detecting a first image and a second image of the same face; (b) dividing the detected first image and the second image into a predetermined size, respectively, and extracting the first partial images corresponding to the partial images of the divided first image and the partial images of the divided second image. Detecting respective corresponding second partial images; (c) generating first texture information corresponding to texture information of each of the detected first partial images and second texture information corresponding to texture information of each of the detected second partial images; checking similarity of each of the first texture information and the second texture information corresponding to the first texture information; And (e) detecting, according to the examined similarity, first classifiers capable of recognizing the identity of the face from the first partial images, the classifier detection having the texture information of the face. Way. The method of claim 1, wherein step (a) And a first image and a second image in a direction in which the face faces the front. The classifier detection method having texture information of the face. The method of claim 1, wherein step (a) And a Gaussian low pass filter for filtering the first image and the second image. The method of claim 1, wherein step (b) And the first partial images are overlapped by a predetermined portion, and the second partial images are overlapped by a predetermined portion, respectively. The method of claim 1, wherein step (c) (c1) detecting the first texture information and the second texture information from the first partial images and the second partial images, respectively, using a local binary pattern (LBP) method; And and (c2) histogramting the detected first texture information and second texture information, respectively. The method of claim 5, wherein step (c1) Haar morphology wavelet method, median morphology wavelet method, Erodent morphology wavelet method and expanded morphology wavelet method among the local binary pattern methods A classifier detection method having texture information of a face, using any one of the methods. The method of claim 1, wherein step (d) Texture of the face characterized in that the similarity is inspected using any one of the chi square distance, the Kullback-Leibler distance and the Jensen-Shannon distance. Classifier detection method with information. The method of claim 1, wherein the classifier detection method having the face texture information comprises: And after step (a), varying the sizes of the detected images, and repeating steps (b) to (e). The method of claim 1, wherein the classifier detection method having the texture information of the face comprises: and (f) after step (e), detecting second classifiers of the detected first classifiers useful for recognizing the identity of the face using a Bayesian network method. A classifier detection method having texture information of a face. (a) detecting an image of a face; (b) detecting partial images corresponding to classifiers useful for recognizing the identity of the face using the Bayesian Network method from the detected image; (c) generating texture information of each of the detected partial images; (d) checking the similarity between the generated texture information and texture information of the face previously stored; And and (e) recognizing whether or not the face is identical according to the checked similarity. The method of claim 10, wherein step (a) Face recognition method using the statistical characteristics of the texture information, characterized in that for detecting the image of the face facing the front. The method of claim 10, wherein step (a) And a Gaussian low pass filter for filtering the image of the face. delete The method of claim 10, wherein step (b) Face recognition method using the statistical characteristics of the texture information, characterized in that each of the partial images overlap by a predetermined portion. The method of claim 10, wherein step (c) (c1) detecting the texture information from each of the separated images using a local binary pattern (LBP) method; And (c2) a histogram of the detected texture information, characterized in that the face recognition method using the statistical characteristics of the texture information. The method of claim 15, wherein step (c1) Haar morphology wavelet method, median morphology wavelet method, Erodent morphology wavelet method and expanded morphology wavelet method among the local binary pattern methods Face recognition method using the statistical characteristics of the texture information, characterized in that using any one of the methods. The method of claim 10, wherein step (d) The similarity is checked using any one of the following measures: chi square distance, Kullback-Leibler distance and Jensen-Shannon distance. Face Recognition Method using Statistical Characteristics. A face image detector detecting a first image or a second image of the same face; The detected first image or the second image is divided into a predetermined size, and the first partial images corresponding to the partial images of the divided first image or the second images corresponding to the partial images of the divided second image. A partial image detector for detecting partial images; A texture information generator configured to generate first texture information corresponding to texture information of each of the detected first partial images or second texture information corresponding to texture information of each of the detected second partial images; A similarity checker that checks similarity between each of the first texture information and the second texture information corresponding to the first texture information; And A classifier having face information, characterized in that it comprises a first classifier detector for detecting first classifiers capable of recognizing the identity of the face from the first partial images according to the examined similarity. Detection device. The apparatus of claim 18, wherein the face image detection unit And a first image or a second image in a direction in which the face faces the front. The classifier detection device having texture information of the face. The apparatus of claim 18, wherein the face image detection unit And a Gaussian low pass filter for filtering the first image or the second image using a Gaussian low pass filter. 19. The method of claim 18, wherein the partial image detection unit And detecting the images such that the first partial images overlap each other by a predetermined portion, or detecting the images so that the second partial images overlap each other by a predetermined portion. 19. The apparatus of claim 18, wherein the texture information generator An information detector for detecting the first texture information from the first partial images or the second texture information from the second partial images using a local binary pattern (LBP) method; And And a histogram unit configured to histogram the detected first texture information or second texture information. The method of claim 22, wherein the information detection unit Haar morphology wavelet method, median morphology wavelet method, Erodent morphology wavelet method and expanded morphology wavelet method among the local binary pattern methods A classifier detection device having texture information of a face, using any one of the methods. The method of claim 18, wherein the similarity inspection unit Texture of the face characterized in that the similarity is inspected using any one of the chi square distance, the Kullback-Leibler distance and the Jensen-Shannon distance. Classifier detection device having information. 19. The apparatus of claim 18, wherein the classifier detection device having texture information of the face comprises: And a second classifier detector configured to detect second classifiers useful for recognizing the identity of the face among the detected first classifiers using a Bayesian network method. Classifier detection device having a. A face image detector detecting an image of a face; A partial image detector for detecting partial images corresponding to classifiers useful for recognizing the identity of the face from the detected image; A texture information generator configured to generate texture information of each of the detected partial images; A similarity checker that checks the similarity between the generated texture information and texture information of the face previously stored; And And a face recognition unit configured to recognize whether or not the face is identical according to the checked similarity. The apparatus of claim 26, wherein the face image detection unit Facial recognition apparatus using the statistical characteristics of the texture information, characterized in that for detecting the image of the face facing the front. The apparatus of claim 26, wherein the face image detection unit And a Gaussian low pass filter for filtering the image of the face. The method of claim 26, wherein the partial image detection unit A face recognition apparatus using statistical characteristics of texture information, characterized by detecting classifiers detected by a Bayesian network method as the partial images. The method of claim 26, wherein the partial image detection unit And detecting images such that the partial images overlap each other by a predetermined portion. 27. The apparatus of claim 26, wherein the texture information generator An information detector for detecting the texture information from each of the partial images by using a local binary pattern (LBP) method; And And a histogram unit configured to histogram the detected texture information. 32. The apparatus of claim 31, wherein the information detection unit Haar morphology wavelet method, median morphology wavelet method, Erodent morphology wavelet method and expanded morphology wavelet method among the local binary pattern methods Face recognition apparatus using the statistical characteristics of the texture information, characterized in that using any one of the methods. The method of claim 26, wherein the similarity inspection unit The similarity is checked using any one of the following measures: chi square distance, Kullback-Leibler distance and Jensen-Shannon distance. Face Recognition Device using Statistical Characteristics.

Images