KR102357340B1 - Method and apparatus for face recognition - Google Patents

Abstract

A face recognition method and apparatus are disclosed. The face recognition apparatus may obtain a 2D input image including the user's face region, and detect feature points of the face from the 2D input image. The face recognition apparatus may adjust a pose of a pre-stored 3D face model based on the detected feature points, and may generate a 2D projection image from the adjusted 3D face model. The face recognition apparatus may perform face recognition by comparing the face region shown in the 2D input image with the face region shown in the 2D projection image, and output a face recognition result.

Description

FACE RECOGNITION METHOD AND APPARATUS FOR FACE RECOGNITION

The description below relates to a face recognition technology that identifies a user's face.

Face recognition technology, one of the biometric technologies, is evaluated as a convenient and competitive biometric recognition technology because of the advantage of being able to identify the subject in a non-contact manner, unlike recognition technologies that require special actions or actions of the user, such as fingerprint recognition and iris recognition. have. Recently, because of convenience and efficiency, face recognition technology has been actively applied to various application fields such as security systems, mobile authentication, and multimedia data search.

A face recognition method according to an embodiment includes: detecting feature points of a face from a 2D input image; adjusting a pre-stored 3D face model based on the detected feature points; generating a 2D projection image from the adjusted 3D face model; and comparing the 2D input image with the 2D projection image to perform face recognition.

In the face recognition method according to an embodiment, the adjusting of the 3D face model may include adjusting a pose and an expression of the 3D face model by mapping the detected feature points to the 3D face model.

In the face recognition method according to an embodiment, the 3D face model may include a 3D shape model and a 3D texture model, and the 3D shape model and the 3D texture model have a fixed shape of the 3D model, and pose And it may be a model in which the expression can be changed.

In the face recognition method according to an embodiment, the adjusting of the 3D face model may include: adjusting the 3D shape model based on feature points detected from the 2D input image; and adjusting the 3D texture model based on parameter information of the adjusted 3D shape model.

In the face recognition method according to an embodiment, the adjusting of the 3D shape model may include adjusting a pose parameter and an expression parameter of the 3D shape model based on the detected feature points.

In the face recognition method according to an embodiment, generating the 2D projection image may include generating a 2D projection image from the adjusted 3D texture model.

In the face recognition method according to an embodiment, the pre-stored 3D face model may be generated based on feature points detected from a plurality of 2D face images, and the 2D face images include a user's face from a plurality of viewpoints. They may be videos taken in the fields.

In the face recognition method according to an embodiment, the performing of the face recognition may include: determining a degree of similarity between the 2D input image and the 2D projection image; and outputting a face recognition result based on whether the degree of similarity satisfies a predetermined condition.

In the face recognition method according to an embodiment, the detecting of the feature point of the face may include: extracting a face region from the 2D input image; and detecting a feature point included in at least one of eyebrows, eyes, nose, mouth, chin, ears, and an outline of a face from the face region.

A method of generating a 3D face model according to an embodiment includes: acquiring 2D face images of a user photographed at a plurality of viewpoints; detecting feature points of a face from the 2D face images; generating a deformable 3D shape model and a 3D texture model based on the detected feature points; and registering the 3D shape model and the 3D texture model as the user's 3D face model, and storing the 3D face model.

In the method of generating a 3D face model according to an embodiment, the generating of the 3D shape model and the 3D texture model includes: the 3D texture based on the 3D shape model and texture information extracted from at least one of the 2D face images. It may include generating a model.

In the method for generating a 3D face model according to an embodiment, the generating of the 3D shape model and the 3D texture model may include: determining a parameter for mapping the detected feature points and feature points of the 3D standard model; and generating the 3D shape model by applying the determined parameter to the 3D standard model.

According to another embodiment, a method for generating a 3D face model includes: acquiring 2D face images obtained by photographing a user's face and direction data of the 2D face images; determining correspondence point information between the 2D face images; generating 3D data about the user's face based on direction data of the 2D face images and information on correspondence points between the 2D face images; and transforming a 3D standard model into a 3D face model for the user by using the 3D data.

In the method for generating a 3D face model according to another embodiment, the obtaining of direction data of the 2D face images may include obtaining direction data of the 2D face images by using motion data sensed through a motion sensor.

In the method of generating a 3D face model according to another embodiment, the 3D data on the user's face may be a set of 3D points constituting the shape of the user's face.

In the method for generating a 3D face model according to another embodiment, the transforming the 3D standard model into a 3D face model for the user may include matching the 3D standard model with 3D data about the user's face to match the 3D standard model to the 3D data about the user's face. may be transformed into a 3D face model for the user.

A face recognition apparatus according to an embodiment includes: an image acquisition unit configured to acquire a 2D input image including a user's face region; a 3D face model processing unit that adjusts a pose of a pre-stored 3D face model according to a pose of a user's face displayed in the 2D input image, and generates a 2D projection image from the 3D face model; and a face recognition unit configured to perform face recognition by comparing the 2D input image with the 2D projected image.

In the face recognition apparatus according to an embodiment, the 3D face model processor may include: a face region detector configured to detect a face region from the 2D input image; and a feature point detector configured to detect a feature point of a face from the face region.

In the face recognition apparatus according to an embodiment, the 3D face model processing unit may adjust the pose of the 3D face model by matching the detected feature points with the feature points of the 3D face model stored in advance.

In the face recognition apparatus according to an embodiment, the 3D face model processing unit adjusts a pose of a 3D shape model according to a face pose of a user displayed in the 2D input image, and based on parameter information of the adjusted 3D shape model You can adjust the 3D texture model.

The face recognition apparatus according to an embodiment may further include a display configured to display at least one of the 2D input image, the 2D projected image, and the face recognition result.

An apparatus for generating a 3D face model according to an embodiment includes: an image acquisition unit configured to acquire 2D face images of a user photographed at a plurality of viewpoints; a feature point detector for detecting feature points of a face from the 2D face images; a 3D face model generator that generates a deformable 3D shape model and a 3D texture model based on the detected feature points; and a 3D face model registration unit for registering and storing the 3D shape model and the 3D texture model as the user's 3D face model.

In an apparatus for generating a 3D face model according to an embodiment, the 3D face model generator may include: a 3D shape model generator configured to generate a 3D shape model of the user's face based on the detected facial feature points; and a 3D texture model generator configured to generate a 3D texture model based on texture information extracted from at least one of the 2D face images and the 3D shape model.

According to another embodiment, an apparatus for generating a 3D face model includes an image acquisition unit configured to acquire 2D face images of a user photographed at a plurality of viewpoints; a motion sensing unit to obtain direction data of the 2D face images; 3D data about the user's face is generated based on correspondence point information between the 2D face images and direction data of the 2D face images, and a 3D standard model is transformed into a 3D face model for the user using the 3D data 3D face model generation unit; and a 3D face model registration unit for registering and storing a 3D face model for the user.

1 is a diagram for explaining an overall operation of a face recognition system according to an exemplary embodiment.
2 is a diagram illustrating a detailed configuration of an apparatus for generating a 3D face model according to an exemplary embodiment.
3 is a diagram illustrating a detailed configuration of a face recognition apparatus according to an exemplary embodiment.
4 is a diagram for describing a process of detecting a feature point from a plurality of 2D face images according to an exemplary embodiment.
5 is a diagram for explaining a process of generating a 3D face model using a 3D standard model according to an exemplary embodiment.
6 is a diagram for explaining a process of adjusting a 3D face model based on feature points detected in a 2D input image according to an exemplary embodiment.
7 is a diagram for explaining a process of performing face recognition by comparing a 2D input image with a 2D projection image, according to an exemplary embodiment.
8 is a flowchart illustrating an operation of a method for generating a 3D face model according to an exemplary embodiment.
9 is a flowchart illustrating an operation of a face recognition method according to an exemplary embodiment.
10 is a diagram illustrating a detailed configuration of an apparatus for generating a 3D face model according to another exemplary embodiment.
11 is a flowchart illustrating an operation of a method for generating a 3D face model according to another exemplary embodiment.

Hereinafter, embodiments will be described in detail with reference to the accompanying drawings. The specific structural and functional descriptions below are only exemplified for the purpose of describing the embodiments, and the scope of the embodiments should not be construed as being limited to the content described in the text. Various modifications and variations are possible from these descriptions by those skilled in the art. In addition, the same reference numerals shown in each drawing indicate the same members, and well-known functions and structures will be omitted.

1 is a diagram for explaining an overall operation of a face recognition system according to an exemplary embodiment. The face recognition system 100 may recognize a user's face from a 2D input image for face recognition. The face recognition system 100 may analyze the 2D input image to extract and identify the user's face displayed in the 2D input image. The face recognition system 100 may be utilized in various application areas, such as a monitoring/security system, mobile authentication, or multimedia data search.

The face recognition system 100 may register a 3D (dimensional) face model for the user and perform face recognition using the registered 3D face model. The 3D face model is a deformable 3D model and may be deformed according to a user's face pose or facial expression displayed in the 2D input image. For example, when the face pose shown in the 2D input image is a pose facing the left side, the face recognition system 100 may rotate the registered 3D face model to face the left side. Also, the face recognition system 100 may adjust the expression of the 3D face model according to the expression of the user's face displayed in the 2D input image. For example, the face recognition system 100 analyzes the user's facial expression based on the feature points of the face detected from the 2D input image, and the shape of the eyes, mouth, nose, etc. of the 3D face model to correspond to the analyzed facial expression. can be adjusted.

The face recognition system 100 may generate a 2D projection image from a registered 3D face model and perform face recognition by comparing the 2D projection image with the 2D input image. Face recognition may be performed in real time by using 2D images. A 2D projection image is a 2D image displayed by projecting a 3D face model onto a plane. For example, a 2D projection image is a 2D image in which a 3D face model matched with a 2D input image is projected from the same viewpoint as a viewpoint in the 2D input image. It may be the same as the pose of the user's face shown in the image. Face recognition may be performed by matching a pre-stored 3D face model to a face pose shown in the 2D input image, and performing face recognition by comparing the 2D projection image with the 2D input image. Even if the user's face pose shown in the 2D input image is not a frontal pose, the recognition rate according to the pose change may be improved by performing face recognition by matching the 3D face model to the face pose shown in the 2D input image.

Hereinafter, the operation of the face recognition system 100 described above will be described in more detail. The face recognition operation performed by the face recognition system 100 largely includes a process of registering a user's 3D face model (110) and a process of recognizing a user's face from a 2D input image using the registered 3D face model (120). ) may be included.

In the process 110 of registering the user's 3D face model, the face recognition system 100 may acquire 130 a plurality of 2D face images of the same user for face registration. The plurality of 2D face images may include images obtained by photographing the user's face from various viewpoints. For example, the face recognition system 100 may acquire 2D face images obtained by photographing the user's face from the front and the side through the camera. The 2D face image represents an image including the user's face region, but it is not necessary to include the entire face region. The face recognition system 100 may detect 140 a feature point (or landmark) of a face from a plurality of obtained 2D face images. For example, the face recognition system 100 may detect a feature point including an eyebrow, eye, nose, mouth, chin, hair, ear, and/or an outline of a face from the 2D face images.

The face recognition system 100 may personalize the 3D model by applying feature points extracted from 2D face images for face registration to a predetermined 3D standard model ( 150 ). The 3D standard model may be, for example, a deformable 3D shape model generated based on 3D face learning data. The 3D standard model may be composed of a 3D shape and a texture, and may include parameters representing the 3D shape. The face recognition system 100 may generate a 3D face model of the user's face by matching feature points of the 3D standard model with feature points extracted from 2D face images. The generated 3D face model may be registered and stored as a 3D face model of the user displayed in the 2D face images.

According to another embodiment, the face recognition system 100 may generate a 3D face model for the user using a plurality of 2D face images for face registration, motion data for the 2D face images, and a 3D standard model. . The face recognition system 100 may acquire direction data of 2D face images through a motion sensor together with the 2D face images, and may relate to the user's face based on matching information and direction data of the 2D face images. 3D data can be created. The 3D data about the user's face may be a set of 3D points constituting the shape of the user's face. The face recognition system 100 may generate a 3D face model of the user by matching 3D data regarding the user's face to the 3D standard model. The generated 3D face model may be stored and registered as a 3D face model of a user displayed in 2D face images.

In the process 120 of recognizing the user's face from the 2D input image, the face recognition system 100 may obtain a 2D input image including the user's face region through the camera. The face recognition system 100 may perform face recognition with only one 2D input image, but the scope of the embodiment is not limited thereto. The face recognition system 100 may adjust 160 the pre-stored 3D face model according to the face pose or expression displayed in the 2D input image. The face recognition system 100 may adjust the pose of the 3D face model to match the facial pose shown in the 2D input image, and adjust the expression of the 3D face model to match the facial expression shown in the 2D input image.

The face recognition system 100 may generate a 2D projection image from a 3D face model matched with a 2D input image for face recognition. The face recognition system 100 may perform face recognition by comparing the 2D input image and the 2D projection image with each other ( 170 ), and may output a face recognition result. For example, the face recognition system 100 determines the similarity between the face region shown in the 2D input image and the face region shown in the 2D projection image, and when the similarity satisfies a predetermined condition, the recognition result of “successful face recognition” may be output, otherwise, a recognition result of “face recognition failure” may be output.

The face recognition system 100 may include any one of the 3D face model generating apparatus 200 of FIG. 2 and the 3D face model generating apparatus 1000 of FIG. 10 , and the face recognition apparatus 300 of FIG. 3 . The process 110 of registering the user's 3D face model may be performed by the face model generating apparatus 200 of FIG. 2 or the face model generating apparatus 1000 of FIG. 10 . The process 120 of recognizing the user's face from the 2D input image may be performed by the face recognition apparatus 300 .

2 is a diagram illustrating a configuration of an apparatus for generating a 3D face model according to an exemplary embodiment. The 3D face model generating apparatus 200 may generate a 3D face model of the user's face from a plurality of 2D face images for face registration. The 3D face model generating apparatus 200 may generate a 3D shape model and a 3D texture model as a 3D face model, and register the generated 3D shape model and 3D texture model as a 3D face model for the user. Referring to FIG. 2 , the 3D face model generating apparatus 200 may include an image obtaining unit 210 , a feature point detecting unit 220 , a 3D face model generating unit 230 , and a 3D face model registering unit 260 .

The image acquisition unit 210 may acquire 2D face images of the user for face registration. The 2D face images may include a user's face region photographed in various face poses. The image acquisition unit 210 may acquire 2D face images captured at a plurality of viewpoints, such as a front image and a side image, through a camera. Overall 2D shape information and texture information of the user's face may be extracted from the front image, and more specific information about the face shape may be extracted from the side image. For example, information about the three-dimensional shape of the user's face may be determined through comparison processing between the user's face regions displayed in the front image and the side image. According to an embodiment, the user may photograph a plurality of 2D face images through a camera in order to register the 3D face model, and the image acquisition unit 210 may store the 2D face images photographed through the camera.

The feature point detector 220 may detect a face region from the 2D face image, and detect feature points (or landmarks) of the face within the detected face region. For example, the feature point detector 220 may detect a feature point located in an outline such as eyebrows, eyes, nose, mouth, and/or chin in the 2D face images. According to an embodiment, the feature point detection unit 220 uses an active shape model (ASM), an active appearance model (AAM), or a supervised descent method (SDM) from 2D face images. Facial feature points can be detected.

The 3D face model generator 230 may generate a 3D face model of the user's face based on feature points detected from the 2D face images. As the 3D face model, a 3D shape model and a 3D texture model that can be deformed for a user's face may be generated. The 3D face model generator 230 may include a 3D shape model generator 240 and a 3D texture model generator 250 .

The 3D shape model generator 240 may generate a 3D shape model of the user's face by using a plurality of 2D face images captured at different viewpoints. A 3D shape model is a 3D model that has a shape but no texture. The 3D shape model generator 240 may generate a 3D shape model based on feature points of the face detected from the 2D face images. The 3D shape model generator 240 may determine a parameter for mapping the feature points detected from the 2D face images and the feature points of the 3D standard model, and apply the determined parameter to the 3D standard model to generate the 3D shape model. . For example, the 3D shape model generating unit 240 matches the feature points of the 3D standard model with feature points, such as eyebrows, eyes, nose, mouth, and/or chin, detected in 2D face images, to the user's face. You can create a 3D shape model for

By generating a 3D shape model using a 2D face image captured at a different viewpoint, a more sophisticated 3D shape model may be generated. When a 3D shape model is generated using only a frontal image of a user's face taken from the front, it may be difficult to determine a three-dimensional shape such as the height of the nose and the shape of the cheeks from the 3D shape model. However, when a 3D shape model is generated using a plurality of 2D face images taken from different viewpoints, information on the height of the nose and the shape of the cheeks can be additionally reflected, so that the 3D shape model can be formed more precisely. have.

The 3D texture model generator 250 may generate a 3D texture model based on the texture information extracted from at least one of the 2D face images and the 3D shape model. For example, the 3D texture model generator 250 may generate a 3D texture model by mapping the texture extracted from the front image to the 3D shape model. The 3D texture model is a model in which both the shape and texture of the 3D model exist. The 3D texture model has a higher level of detail than the 3D shape model, and may include vertices of the 3D shape model. In the 3D shape model and the 3D texture model, the shape of the 3D model is fixed, and poses and expressions can be changed. The 3D shape model and the 3D texture model may represent the same poses and expressions by the same parameters.

The 3D face model registration unit 260 may register and store the 3D shape model and the 3D texture model as the user's 3D face model. For example, if the user of the 2D face image acquired by the image acquisition unit 210 says “A”, the 3D face model registration unit 350 sets the 3D shape model and 3D texture model generated for “A” to “A”. It is possible to register as a 3D face model of ", and store the 3D shape model and 3D texture model of "A" in the 3D face model storage unit (not shown).

3 is a diagram illustrating a detailed configuration of a face recognition apparatus according to an exemplary embodiment. The face recognition apparatus 300 may perform face recognition on a user displayed in a 2D input image for face recognition by using the registered 3D face model. The face recognition apparatus 300 may generate a 2D projection image after rotating the 3D face model to the same pose as the user's face pose shown in the 2D input image. The face recognition apparatus 300 may perform face recognition by comparing the 2D projection image with the 2D input image. The face recognition apparatus 300 performs face recognition by matching a registered 3D face model to a face pose displayed in a 2D input image, thereby providing a face recognition method robust to a change in a user's pose. Referring to FIG. 3 , the face recognition apparatus 300 may include an image acquisition unit 310 , a 3D face model processing unit 320 , and a face recognition unit 350 . The 3D face model processing unit 320 may include a face region detection unit 330 and a feature point detection unit 340 .

The image acquisition unit 310 may acquire a 2D input image for face recognition including the user's face region. The image acquisition unit 310 may acquire a 2D input image for user recognition or user authentication through a camera or the like. The face recognition apparatus 300 may perform face recognition for a user using one 2D input image, but the scope of the embodiment is not limited thereto.

The face region detector 330 may detect a face region from the 2D input image. The face region detector 330 may identify a face region in the 2D input image using information such as brightness distribution, object movement, color distribution, or eye position in the 2D input image, and extract location information of the face region. . For example, the face region detector 330 may detect the face region from the 2D input image by using a Haar-like-based Adaboost classifier generally used in the related art.

The feature point detector 340 may detect the feature point of the face from the face region in the 2D input image. For example, the feature point detector 340 may detect a feature point including the outline of the eyebrows, eyes, nose, mouth, chin, hair, ears, and/or face from the face region. According to an embodiment, the feature point detector 340 may detect the feature point of the face from the 2D input image using an Active Shape Model, an Active Appearance Model, or a Supervised Descent Method.

The 3D face model processing unit 320 may adjust a pre-stored 3D face model based on the detected feature points. The 3D face model processing unit 320 may match the 3D face model to the 2D input image based on the detected feature points. As a result of the matching, the 3D face model may be deformed to match the facial pose and expression displayed in the 2D input image. The 3D face model processing unit 320 may adjust the pose and expression of the 3D face model by mapping the feature points detected from the 2D input image to the 3D face model. The 3D face model may include a 3D shape model and a 3D texture model. The 3D shape model can be used to quickly match the facial poses shown in the 2D input image, and the 3D texture model can be used to generate a high-resolution 2D projection image.

The 3D face model processing unit 320 may adjust the pose of the 3D shape model according to the pose shown in the 2D input image. The 3D face model processing unit 320 may match the pose of the 3D shape model to the pose shown in the 2D input image by matching the feature points detected in the 2D input image with the feature points of the 3D shape model. The 3D face model processing unit 320 may adjust a pose parameter and an expression parameter of the 3D shape model based on the feature points detected from the 2D input image.

Also, the 3D face model processing unit 320 may adjust the 3D texture model based on parameter information of the 3D shape model. The 3D face model processing unit 320 may apply the pose parameter and the expression parameter determined in the process of matching the 2D input image and the 3D shape model to the 3D texture model. As a result of the application, the 3D texture model may be adjusted to have the same pose and expression as that of the 3D shape model. After adjusting the 3D texture model, the 3D face model processing unit 320 may generate a 2D projection image by projecting the adjusted 3D texture model onto a plane.

The face recognition unit 350 may perform face recognition by comparing the 2D input image with the 2D projection image. The face recognizer 350 may perform face recognition based on the similarity between the face region shown in the 2D input image and the face region shown in the 2D projection image. The face recognition unit 350 may determine a similarity between the 2D input image and the 2D projected image, and output a face recognition result based on whether the similarity satisfies a predetermined condition.

The face recognition unit 350 may use a feature value determination method generally used in the field of face recognition technology to determine the similarity between the 2D input image and the 2D projection image. For example, the face recognition unit 350 uses a filter for feature extraction such as a Gabor filter, a local binary pattern (LBP), a histogram of oriented gradient (HOG), a principal component analysis (PCA), or a linear discriminant analysis (LDA). The similarity between the 2D input image and the 2D projection image may be determined using The Gabor filter is a filter that extracts features from an image using multiple filters having various sizes and angles, and the LBP is a filter that extracts features from the difference between the current pixel and neighboring pixels in the image. According to an embodiment, the face recognition unit 350 divides the face region shown in the 2D input image and the 2D projection image into cells of a certain size, and then a histogram of the LBP for each cell (the LBP belonging to the corresponding cell). Histograms of the index values) can be calculated. The face recognition unit 350 determines a vector obtained by connecting the calculated histograms in a line as a final feature value, compares the final feature value of the 2D input image with the final feature value of the 2D projection image, and then compares the final feature value between the 2D input image and the 2D projection image. similarity can be determined.

According to an embodiment, the face recognition apparatus 300 may further include a display unit 360 . The display unit 360 may display a 2D input image, a 2D projection image, and/or a face recognition result. When the user determines that the face is not photographed properly based on the displayed 2D input image, or when it is finally displayed that the face recognition has failed, the user may photograph the face again, and the face recognition device 300 is photographed again. Face recognition may be performed again on the generated 2D input image.

4 is a diagram for describing a process of detecting a feature point from a plurality of 2D face images according to an exemplary embodiment. Referring to FIG. 4 , an image 420 represents a 2D face image obtained by photographing a user's face from the front, and images 410 and 430 represent 2D face images obtained by photographing a user's face from the side. . Overall 2D shape information and texture information of the user's face may be extracted from the image 420 taken from the front. More specific information about the face shape may be extracted from the images 410 and 430 taken from the side. For example, a basic model for the user's face is set based on the user's face shape extracted from the image 420 , and a three-dimensional model of the corresponding basic model is set based on the user's face shape extracted from the images 410 and 430 . shape can be determined.

The feature point detector 220 of the 3D face model generating apparatus 200 may detect feature points of a face from 2D face images captured at a plurality of viewpoints, such as images 410 , 420 , and 430 . The facial feature points represent feature points located in contour regions such as eyebrows, eyes, nose, mouth, and chin. The feature point detector 220 may detect the feature point of the face from the images 410 , 420 , and 430 using a method such as ASM, AAM, or SDM, which is generally known in the related art. Initialization of the pose, scale, or position of the ASM model (or the AAM model or the SDM model) may be performed based on a face detection result.

The image 440 represents an image as a result of detecting feature points 444 in the face region 442 of the image 410 , and the image 450 shows a feature point 454 in the face region 452 of the image 420 . The detected result image is shown. Similarly, the image 460 represents an image as a result of detecting feature points 464 in the face region 462 of the image 430 .

5 is a diagram for explaining a process of generating a 3D face model using a 3D standard model according to an exemplary embodiment. Referring to FIG. 5 , a model 510 represents a 3D standard model. The 3D standard model is, for example, a deformable 3D shape model generated based on 3D face learning data, and may be a parametric model capable of representing the identity of a user's face through an average shape and parameters. can

The 3D standard model may be composed of an average shape and a shape change amount as shown in Equation 1 below. The shape change amount is a weighted sum of the shape parameter and the shape vector.

는 3D 표준 모델의 3D 형태를 구성하는 엘리먼트들을 나타내고,

는 3D 표준 모델의 평균 형태에 관한 엘리먼트들을 나타낸다.

는 인덱스 계수 i에 대응하는 형태 엘리먼트들을 나타내고,

는 인덱스 계수 i에 대응하는 형태 엘리먼트에 적용되는 형태 파라미터를 나타낸다.here,

represents the elements constituting the 3D shape of the 3D standard model,

denotes elements related to the average shape of the 3D standard model.

represents the form elements corresponding to the index coefficient i,

denotes a shape parameter applied to a shape element corresponding to the index coefficient i.

는 다음의 수학식 2와 같이 3차원 점들의 좌표로 구성될 수 있다.Elements that make up the 3D shape of the 3D standard model

may be composed of coordinates of three-dimensional points as shown in Equation 2 below.

는

는 3차원 점(

,

,

)의 인덱스를 나타내는 변수이고, T는 트랜스포즈(transpose)를 나타낸다.here,

Is

is a three-dimensional point (

,

,

) is a variable indicating the index, and T indicates transpose.

In order to register the user's face, the 3D shape model generator 240 of the 3D face model generating apparatus 200 may personalize the 3D standard model based on 2D face images captured at a plurality of viewpoints. The 3D shape model generator 240 determines a parameter capable of matching the feature points included in the 3D standard model with the feature points of the face detected from the 2D face images, and applies the determined parameter to the 3D standard model to apply to the user's face. You can create a 3D shape model for

The models 520 and 530 represent 3D shape models of the user's face generated from the 3D standard model 510 . The model 520 represents a 3D shape model viewed from the front, and the model 530 represents a 3D shape model viewed from the side. The 3D shape model is a model in which shape information exists but texture information does not exist, and can be used for high-speed matching with a 2D input image in a user authentication process.

The 3D texture model generator 250 may generate a 3D texture model by mapping a texture extracted from at least one 2D face image among the 2D face images to the surface of the 3D shape model. Mapping the texture to the surface of the 3D shape model may indicate, for example, adding depth information obtained from the 3D shape model to texture information extracted from a 2D face image photographed from the front.

The models 540 and 550 represent 3D texture models generated based on the 3D shape model. The model 540 represents a 3D texture model viewed from the front, and the model 550 represents a 3D texture model viewed from a diagonal direction. The 3D texture model is a model in which both shape information and texture information exist, and may be used to generate a 2D projection image in a user authentication process.

The 3D shape model and the 3D texture model are 3D models in which the shape of the face representing the user's unique characteristics is fixed and the pose or expression can be changed. The 3D texture model is more precise than the 3D shape model and contains more vertices. Vertices constituting the 3D shape model are a subset of vertices constituting the 3D texture model, and the 3D shape model and the 3D texture model may exhibit the same pose and expression by the same parameters.

6 is a diagram for describing a process of adjusting a 3D face model based on feature points detected in a 2D input image according to an exemplary embodiment. Referring to FIG. 6 , an image 610 is a 2D input image input to a face recognition device for face recognition or user authentication, and represents a face pose image captured by a camera.

The face region detector 330 of the face recognition apparatus 300 detects a face region from the 2D input image, and the feature point detector 340 detects the contours of eyes, eyebrows, nose, mouth, or chin within the detected face region. Positioned feature points can be detected. For example, the feature point detector 340 may detect the feature point of the face from the 2D input image using a method such as ASM, AAM, or SDM. The image 620 is a result image in which the face region 630 is detected from the image 610 by the face region detector 330 and feature points 640 are detected in the face region 630 by the feature point detector 340 . indicates

The 3D face model processing unit 320 may match the previously registered and stored 3D shape model to the 2D input image. The 3D face model processing unit 320 may adjust poses and expressions by adjusting parameters of the 3D shape model based on feature points of the face detected in the 2D input image. The model 650 represents a 3D model of the user's face that has been registered and stored in advance, and the model 660 is a 3D model in which poses and expressions are adjusted based on feature points 640 of the face detected in the image 610 . indicates The 3D face model processing unit 320 may adjust the pose of the 3D shape model stored in advance to be the same as the pose of the user's face displayed in the 2D input image. In the image 610 that is the 2D input image, the user is taking a pose in which the face is rotated to the side, and the 3D shape model whose pose is adjusted by the 3D face model processing unit 320 is also the same as the user pose in the image 610 . He is posing with his face turned to the side.

7 is a view for explaining a process of performing face recognition by comparing a 2D input image with a 2D projection image, according to an exemplary embodiment. The 3D face model processing unit 320 of the face recognition apparatus 300 may adjust a pose parameter and an expression parameter of the 3D shape model based on facial feature points detected from a 2D input image for face recognition. The 3D face model processing unit 320 may apply the adjusted pose parameters and expression parameters of the 3D shape model to the 3D texture model to adjust the poses and expressions of the 3D texture model to be identical to those of the 3D shape model. Thereafter, the 3D face model processing unit 320 may generate a 2D projection image by projecting the 3D texture model onto the image plane. The face recognition unit 350 may perform face recognition based on the similarity between the 2D input image and the 2D projection image, and may output a face recognition result.

Referring to FIG. 7 , an image 710 represents a 2D input image used for face recognition. The image 720 is a reference image compared with an image 710 that is a 2D input image in order for the face recognition unit 350 to perform face recognition. The region 730 included in the image 720 represents a region in which the 2D projection image generated from the 3D texture model is reflected. For example, the region 730 may be a face region in which a texture model in which a texture is mapped to the shape model 660 of FIG. 6 is projected onto an image plane. The face recognition unit 350 may perform face recognition by comparing only the face region of the user shown in the 2D input image with the face region shown in the 2D projection image. Alternatively, as shown in FIG. 7 , the face recognition unit 350 performs face recognition by comparing the entire area of an image 710 that is a 2D input image and an image 720 that is a result image in which a 2D projection image is reflected in the 2D input image. may be

8 is a flowchart illustrating an operation of a method for generating a 3D face model according to an exemplary embodiment.

In operation 810 , the apparatus for generating a 3D face model may acquire 2D face images of the user captured from a plurality of viewpoints through a camera. 2D face images may be used to register a user's face. The 2D face images may include images captured with various face poses, such as a front image and a side image, for example.

In operation 820 , the 3D face model generating apparatus may detect feature points of the face from the 2D face images. For example, the 3D face model generating apparatus detects facial feature points located in the contours of eyebrows, eyes, nose, mouth, etc. from 2D face images using ASM, AAM, or SDM, which are generally known in the related art field. can do.

In operation 830, the 3D face model generating apparatus may generate a 3D shape model based on the detected feature points. The 3D face model generating apparatus may generate a 3D shape model by matching feature points such as eyebrows, eyes, nose, mouth, and chin detected in 2D face images with feature points of the 3D standard model. The 3D face model generating apparatus may determine a parameter for mapping feature points detected from the 2D face images and the feature points of the 3D standard model, and apply the determined parameter to the 3D standard model to generate a 3D shape model.

In operation 840, the 3D face model generating apparatus may generate a 3D texture model based on the texture information extracted from the 2D face image and the 3D shape model. The 3D face model generating apparatus may generate a 3D texture model of the user's face by mapping a texture extracted from at least one 2D face image to a 3D shape model. The 3D texture model may exhibit the same poses and facial expressions as the 3D texture model by applying the parameters of the 3D texture model.

In operation 850, the 3D face model generating apparatus may register and store the 3D shape model and the 3D texture model as the user's 3D face model. The stored 3D shape model and 3D texture model may be used to authenticate the user shown in the 2D input image in the user authentication process.

9 is a flowchart illustrating an operation of a face recognition method according to an exemplary embodiment.

In operation 910 , the face recognition apparatus may detect a feature point of a face from a 2D input image for face recognition. The face recognition apparatus may detect a face region from the 2D input image, and detect feature points of a face located in outlines of eyes, eyebrows, nose, chin, mouth, etc. within the detected face region. For example, the face recognition apparatus may detect a face region from a 2D input image using a Haar-like-based Adaboost classifier, and may detect a feature point of a face within the face region using ASM, AAM, or SDM.

In operation 920, the face recognition apparatus may adjust the pre-registered 3D face model of the user based on the feature points detected in the 2D input image. The face recognition apparatus may match a pre-registered 3D face model to the 2D input image based on the feature points detected in the 2D input image. The face recognition apparatus may transform the 3D face model to match the face pose and expression displayed in the 2D input image.

The 3D face model may include a 3D shape model and a 3D texture model, and the face recognition apparatus adjusts the pose of the 3D shape model based on feature points detected in the 2D input image, and the parameter of the 3D shape model with the adjusted pose Based on the information, the 3D texture model can be adjusted. A pose parameter and an expression parameter of the 3D shape model may be adjusted based on feature points detected from the 2D input image of the face recognition apparatus, and the adjusted parameters of the 3D shape model may be applied to the 3D texture model. As a result of the application of the parameters, the 3D texture model may be adjusted to the same pose and expression as that of the 3D shape model.

In operation 930, the face recognition apparatus may generate a 2D projection image from the 3D texture model. The face recognition apparatus may generate a 2D projection image by projecting the 3D texture model adjusted based on the 3D shape model to a plane in operation 920 . The face pose shown in the 2D projection image may be the same as the face pose shown in the 2D input image. For example, if the user's face pose shown in the 2D input image is a pose looking to the side, the 2D projection image generated through steps 910 to 930 also has a 3D texture model in the same way as in the 2D input image. It can represent a face pose looking to the side.

In operation 940 , the face recognition apparatus may perform face recognition by comparing the 2D input image with the 2D projection image. The face recognition apparatus may perform face recognition based on a degree of similarity between the face region shown in the 2D input image and the face region shown in the 2D projection image. The face recognition apparatus may determine a similarity between the 2D input image and the 2D projected image, and output a face recognition result based on whether the similarity satisfies a predetermined condition. For example, if the similarity between the 2D input image and the 2D projection image satisfies a predetermined condition, the face recognition device outputs a recognition result of “face recognition success”, and otherwise, a recognition result of “face recognition failure” can be printed out.

10 is a diagram illustrating a detailed configuration of an apparatus for generating a 3D face model according to another exemplary embodiment. The 3D face model generating apparatus 1000 may generate a 3D face model of the user's face from a plurality of 2D face images for face registration. The 3D face model generating apparatus 1000 may generate a 3D face model for the user by using 2D face images photographed in various directions, motion data for the 2D face images, and a 3D standard model. Referring to FIG. 10 , the 3D face model generating apparatus 1000 may include an image obtaining unit 1010 , a motion sensing unit 1020 , a 3D face model generating unit 1030 , and a 3D face model registering unit 1040 . .

The image acquisition unit 1010 may acquire 2D face images captured at a plurality of viewpoints for face registration. The image acquisition unit 1010 may acquire 2D face images obtained by photographing the user's face in various directions through a camera. For example, the image acquisition unit 1010 may acquire 2D face images captured from a plurality of viewpoints, such as a front image and a side image.

The motion sensing unit 1020 may obtain direction data of 2D face images. The motion sensing unit 1020 may determine direction data of 2D face images by using motion data sensed through various sensors. The direction data of the 2D face images may include information about a direction in which each 2D face image is captured. For example, the motion sensing unit 1020 uses an inertial measurement unit (IMU) including an accelerometer, a gyroscope, and a magnetometer to obtain respective 2D face images. direction data can be determined.

For example, the user may photograph a face by rotating the camera to various positions, and 2D face images photographed at various viewpoints may be obtained as a result of the photographing. When the 2D face images are captured, the motion sensing unit 1020 performs the speed, direction, roll, pitch and/or speed of the camera capturing the 2D face images based on the sensing information output from the inertial measurement device. Motion data such as a change in yaw may be calculated, and direction data regarding a direction in which 2D face images are captured may be determined from the motion data.

The 3D face model generator 1030 may generate a 3D face model of the user shown in the 2D face images. The 3D face model generator 1030 may detect feature points (or landmarks) of the face from the 2D face images. For example, the 3D face model generator 1030 may detect feature points located in contours such as eyebrows, eyes, nose, mouth, and/or chin from the 2D face images. The 3D face model generator 1030 may determine correspondence point information between 2D face images based on facial feature points detected from the 2D face images.

The 3D face model generator 1030 may generate 3D data about the user's face based on feature point information of the face detected from the 2D face images, information on correspondence points between the 2D face images, and direction data of the 2D face images. . The 3D face model generator 1030 may generate 3D data about the user's face using, for example, a conventional stereo matching technique. The 3D data about the user's face may be a set of 3D points constituting the shape or surface of the user's face.

The 3D face model generator 1030 may transform the 3D standard model into a 3D face model of the user by using 3D data about the user's face. The 3D face model generator 1030 may transform the 3D standard model into a 3D face model for the user by matching a deformable 3D standard model with 3D data about the user's face. The 3D face model generator 1030 may transform the 3D standard model into a 3D face model for the user by matching the feature points on the 3D data with the feature points on the 3D standard model. The 3D face model for the user may include a 3D shape model regarding the shape of the user's face and/or a 3D texture model including texture information.

The 3D face model registration unit 1040 may register and store the 3D face model for the user generated by the 3D face model generation unit 1030 . The stored 3D face model of the user is used to recognize the user's face, and the shape may be changed in the face recognition process.

11 is a flowchart illustrating an operation of a method for generating a 3D face model according to another exemplary embodiment.

In operation 1110, the 3D face model generating apparatus may obtain a plurality of 2D face images for face registration and direction data of the 2D face images. The apparatus for generating a 3D face model may acquire 2D face images of a user captured at a plurality of viewpoints through a camera. The 3D face model generating apparatus may acquire 2D face images obtained by photographing the user's face in various directions, such as a front image and a side image.

The 3D face model generating apparatus may obtain direction data of 2D face images by using motion data sensed through a motion sensor. For example, the 3D face model generating apparatus may obtain direction data for each of the 2D face images by using motion data sensed through an inertial measurement device including an acceleration sensor, an angular velocity sensor, and a magnetic sensor. The direction data of the 2D face images may include information about a direction in which each 2D face image is captured.

In operation 1120, the 3D face model generating apparatus may determine correspondence point information between 2D face images. The 3D face model generating apparatus may detect feature points of a face from the 2D face images, and detect corresponding points corresponding to each other between the 2D face images based on the detected feature points.

In operation 1130, the 3D face model generating apparatus may generate 3D data about the user's face. For example, 3D data about the user's face is a set of 3D points constituting the shape of the user's face or the face surface, and may include a plurality of vertices. The 3D face model generating apparatus may generate 3D data about the user's face based on facial feature point information detected from 2D face images, correspondence point information between 2D face images, and direction data of the 2D face images. The 3D face model generating apparatus may generate 3D data about the user's face by using an existing stereo matching technique.

In operation 1140 , the 3D face model generating apparatus may transform the 3D standard model into a 3D face model of the user by using the 3D data generated in operation 1130 . The 3D face model generating apparatus may transform the 3D standard model into a 3D face model of the user by matching the 3D standard model to 3D data regarding the user's face. The 3D face model generating apparatus may generate a 3D face model for the user by matching the feature points on the 3D data with the feature points on the 3D standard model. The 3D face model generating apparatus may generate a 3D shape model and/or a 3D texture model as a 3D face model for the user. The generated 3D face model for the user may be stored and registered, and used to recognize the user's face.

The embodiments described above may be implemented by a hardware component, a software component, and/or a combination of a hardware component and a software component. For example, the apparatus, methods, and components described in the embodiments may include, for example, a processor, a controller, an arithmetic logic unit (ALU), a digital signal processor, a microcomputer, a field programmable gate (FPGA) array), a programmable logic unit (PLU), a microprocessor, or any other device capable of executing and responding to instructions, may be implemented using one or more general purpose or special purpose computers. The processing device may execute an operating system (OS) and one or more software applications running on the operating system. A processing device may also access, store, manipulate, process, and generate data in response to execution of the software. For convenience of understanding, although one processing device is sometimes described as being used, one of ordinary skill in the art will recognize that the processing device includes a plurality of processing elements and/or a plurality of types of processing elements. It can be seen that can include For example, the processing device may include a plurality of processors or one processor and one controller. Other processing configurations are also possible, such as parallel processors.

Software may comprise a computer program, code, instructions, or a combination of one or more thereof, which configures a processing device to operate as desired or is independently or collectively processed You can command the device. The software and/or data may be any kind of machine, component, physical device, virtual equipment, computer storage medium or apparatus, to be interpreted by or to provide instructions or data to the processing device. , or may be permanently or temporarily embody in a transmitted signal wave. The software may be distributed over networked computer systems and stored or executed in a distributed manner. Software and data may be stored in one or more computer-readable recording media.

The method according to the embodiment may be implemented in the form of program instructions that can be executed through various computer means and recorded in a computer-readable medium. The computer-readable medium may include program instructions, data files, data structures, etc. alone or in combination. The program instructions recorded on the computer-readable medium may be specially designed and configured for the embodiment, or may be known and available to those skilled in the art of computer software. Examples of the computer-readable recording medium include magnetic media such as hard disks, floppy disks and magnetic tapes, optical media such as CD-ROMs and DVDs, and magnetic such as floppy disks. - includes magneto-optical media, and hardware devices specially configured to store and execute program instructions, such as ROM, RAM, flash memory, and the like. Examples of program instructions include not only machine language codes such as those generated by a compiler, but also high-level language codes that can be executed by a computer using an interpreter or the like. A hardware device may be configured to operate as one or more software modules to perform the operations of the embodiments, and vice versa.

As described above, although the embodiments have been described with reference to the limited drawings, various modifications and variations are possible by those skilled in the art from the above description. For example, the described techniques are performed in an order different from the described method, and/or the described components of the system, structure, apparatus, circuit, etc. are combined or combined in a different form than the described method, or other components Or substituted or substituted by equivalents may achieve an appropriate result. Therefore, other implementations, other embodiments, and equivalents to the claims are also within the scope of the following claims.

Claims (32)

detecting feature points of a face from a 2D (dimensional) input image;
adjusting a pre-stored 3D face model based on the detected feature points;
generating a 2D projection image from the adjusted 3D face model; and
Comprising the step of performing face recognition by comparing the 2D input image and the 2D projection image,
Adjusting the 3D face model comprises:
adjusting the expression of the 3D face model by mapping the detected feature points to the 3D face model;
The step of generating the 2D projection image comprises:
Generating the 2D projection image from the 3D face model with the expression adjusted,
The facial expression displayed in the generated 2D projection image corresponds to the facial expression displayed in the 2D input image,
How to recognize faces. According to claim 1,
Adjusting the 3D face model comprises:
adjusting the pose of the 3D face model by mapping the detected feature points to the 3D face model
A face recognition method comprising 3. The method of claim 2,
The 3D face model includes a 3D shape model and a 3D texture model,
Adjusting the 3D face model comprises:
adjusting the 3D shape model based on feature points detected from the 2D input image; and
adjusting the 3D texture model based on parameter information of the adjusted 3D shape model
A face recognition method comprising 4. The method of claim 3,
Adjusting the 3D shape model comprises:
adjusting a pose parameter and an expression parameter of the 3D shape model based on the detected feature points
A face recognition method comprising 4. The method of claim 3,
The step of generating the 2D projection image comprises:
generating a 2D projection image from the adjusted 3D texture model;
A face recognition method comprising 4. The method of claim 3,
The 3D shape model and the 3D texture model are,
A face recognition method, in which the shape of a 3D model is fixed, and a model whose poses and expressions can be changed. According to claim 1,
The 2D projection image is
A face recognition method representing the same face pose as the face pose shown in the 2D input image. According to claim 1,
The step of performing the face recognition comprises:
determining a similarity between the 2D input image and the 2D projection image; and
outputting a face recognition result based on whether the degree of similarity satisfies a predetermined condition
A face recognition method comprising According to claim 1,
The step of detecting the feature points of the face comprises:
extracting a face region from the 2D input image; and
detecting feature points included in at least one of eyebrows, eyes, nose, mouth, chin, ears, and face contours from the face region;
A face recognition method comprising acquiring 2D face images of a user taken at a plurality of viewpoints;
detecting feature points of a face from the 2D face images;
generating a deformable 3D shape model and a 3D texture model based on the detected feature points; and
Registering the 3D shape model and the 3D texture model as the user's 3D face model, and storing the model,
In the face recognition process using the registered 3D face model,
In the 3D face model, an expression is adjusted based on feature points detected from a 2D input image for face recognition,
A 2D projection image is generated from the 3D face model with the expression adjusted,
a face recognition result is determined based on the 2D projection image and the 2D input image;
The facial expression shown in the generated 2D projection image corresponds to the facial expression shown in the 2D input image,
How to create a 3D face model. 11. The method of claim 10,
The step of generating the 3D shape model and the 3D texture model comprises:
generating the 3D texture model based on texture information extracted from at least one of the 2D face images and the 3D shape model;
A method of generating a 3D face model, including 11. The method of claim 10,
The step of generating the 3D shape model and the 3D texture model comprises:
determining a parameter for mapping the detected feature points and feature points of a 3D standard model; and
generating the 3D shape model by applying the determined parameter to the 3D standard model
A method of generating a 3D face model, including 11. The method of claim 10,
The 3D texture model is
A method of generating a 3D face model, including vertices of the 3D shape model. 11. The method of claim 10,
The 2D face images are
A method of generating a 3D face model, including images obtained by photographing the user's face from different viewpoints. obtaining 2D face images obtained by photographing a user's face and direction data of the 2D face images;
determining correspondence point information between the 2D face images;
generating 3D data about the user's face based on direction data of the 2D face images and information on correspondence points between the 2D face images; and
transforming a 3D standard model into a 3D face model for the user using the 3D data;
including,
In the face recognition process using the 3D face model for the user,
In the 3D face model, an expression is adjusted based on feature points detected from a 2D input image for face recognition,
A 2D projection image is generated from the 3D face model with the expression adjusted,
a face recognition result is determined based on the 2D projection image and the 2D input image;
The facial expression displayed in the generated 2D projection image corresponds to the facial expression displayed in the 2D input image. 16. The method of claim 15,
Transforming the 3D standard model into a 3D face model for the user comprises:
A method for generating a 3D face model, wherein the 3D standard model is matched with 3D data about the user's face to transform the 3D standard model into a 3D face model for the user. 16. The method of claim 15,
3D data about the user's face,
A method of generating a 3D face model, which is a set of 3D points constituting the shape of the user's face. 16. The method of claim 15,
The step of obtaining direction data of the 2D face images comprises:
A method for generating a 3D face model, comprising obtaining direction data of the 2D face images by using motion data sensed through a motion sensor. 16. The method of claim 15,
The step of determining correspondence point information between the 2D face images includes:
detecting feature points of a face from the 2D face images; and
determining correspondence point information between the 2D face images based on the detected feature points
A method of generating a 3D face model, including A computer-readable recording medium in which a program for executing the method of any one of claims 1 to 19 is recorded. an image acquisition unit for acquiring a 2D input image including a user's face region;
a 3D face model processing unit that adjusts a pre-stored 3D face model based on the feature points detected from the 2D input image and generates a 2D projection image from the adjusted 3D face model; and
A face recognition unit that performs face recognition by comparing the 2D input image with the 2D projection image
including,
The 3D face model processing unit,
a face region detector configured to detect a face region from the 2D input image; and
A feature point detection unit for detecting the feature points of the face from the face region
including,
The 3D face model processing unit,
adjusting the expression of the 3D face model by mapping the detected feature points to the 3D face model, and generating the 2D projection image from the 3D face model with the adjusted expression;
The facial expression displayed in the generated 2D projection image corresponds to the facial expression displayed in the 2D input image. delete 22. The method of claim 21,
The 3D face model processing unit,
and adjusting a pose of the 3D face model according to a pose of a user's face displayed in the 2D input image by matching the detected feature points with the feature points of the 3D face model stored in advance. 24. The method of claim 23,
The 3D face model includes a 3D shape model and a 3D texture model,
The 3D face model processing unit,
A face recognition apparatus for adjusting a pose of the 3D shape model according to a face pose of a user displayed in the input image, and adjusting the 3D texture model based on parameter information of the adjusted 3D shape model. 25. The method of claim 24,
The 3D face model processing unit,
and generating the 2D projection image by projecting the adjusted 3D texture model onto a plane. 22. The method of claim 21,
a display unit configured to display at least one of the 2D input image, the 2D projection image, and a face recognition result
A face recognition device further comprising a. an image acquisition unit for acquiring 2D face images of a user taken at a plurality of viewpoints;
a feature point detector for detecting feature points of a face from the 2D face images;
a 3D face model generator that generates a deformable 3D shape model and a 3D texture model based on the detected feature points; and
3D face model registration unit for registering and storing the 3D shape model and the 3D texture model as the user's 3D face model
including,
In the face recognition process using the registered 3D face model,
In the 3D face model, an expression is adjusted based on feature points detected from a 2D input image for face recognition,
A 2D projection image is generated from the 3D face model with the expression adjusted,
a face recognition result is determined based on the 2D projection image and the 2D input image;
The facial expression displayed in the generated 2D projection image corresponds to the facial expression displayed in the 2D input image. 28. The method of claim 27,
The 3D face model generation unit,
3D face model generating apparatus, which determines a parameter for mapping the detected feature points and feature points of a 3D standard model, and generates the 3D shape model by applying the determined parameter to the 3D standard model. 28. The method of claim 27,
The 3D face model generation unit,
a 3D shape model generator for generating a 3D shape model of the user's face based on the detected facial feature points; and
A 3D texture model generator that generates a 3D texture model based on the texture information extracted from at least one of the 2D face images and the 3D shape model
A 3D face model generating device comprising a. an image acquisition unit for acquiring 2D face images of a user taken at a plurality of viewpoints;
a motion sensing unit to obtain direction data of the 2D face images;
3D data about the user's face is generated based on correspondence point information between the 2D face images and direction data of the 2D face images, and a 3D standard model is transformed into a 3D face model for the user using the 3D data 3D face model generation unit;
3D face model registration unit for registering and storing 3D face models for the user
including,
In the face recognition process using the stored 3D face model,
In the 3D face model, an expression is adjusted based on feature points detected from a 2D input image for face recognition,
A 2D projection image is generated from the 3D face model with the expression adjusted,
a face recognition result is determined based on the 2D projection image and the 2D input image;
The facial expression displayed in the generated 2D projection image corresponds to the facial expression displayed in the 2D input image. 31. The method of claim 30,
The 3D face model generation unit,
and detecting feature points of a face from the 2D face images, and determining correspondence point information between the 2D face images based on the detected feature points. 31. The method of claim 30,
The 3D face model generation unit,
and matching the 3D standard model with 3D data about the user's face to transform the 3D standard model into a 3D face model for the user.

Images