KR20020067088A - Method and apparatus for replacing a model face of moving image - Google Patents

Abstract

PURPOSE: A method and an apparatus for substituting a three-dimensional motion-picture model face are provided to convert an advertising model of motion-picture advertisement information into the face of a specific person. CONSTITUTION: In case that the second face is substituted for the first face of motion-picture information, a three-dimensional personal photo avatar corresponding to a picture including the second face is generated(S800). Face parameters of the first face are extracted from the motion-picture information(S804). The three-dimensional personal photo avatar is deformed according to the extracted face parameters of the first face(S806). The deformed three-dimensional personal photo avatar is substituted for the first face image area in the motion-picture information(S808).

Description

3D moving image model face replacement method and apparatus {METHOD AND APPARATUS FOR REPLACING A MODEL FACE OF MOVING IMAGE}

The present invention relates to a 3D moving image model face replacement apparatus and method, and to a 3D moving image model face replacement apparatus and method for effectively making a face of a 3D moving image model a specific avatar.

With the development of communication technology, the use of the Internet, in which communication terminals worldwide, that is, computers are connected to one communication network, is rapidly increasing.

Through the Internet, users can not only be provided with various information, but also chat or play games with other users connected to the Internet.

In general, users have a desire to become models that appear in various advertisements, music videos, and movies provided through the Internet.

In order to satisfy such a desire, an internet user downloads desired advertisement information, removes the face of the advertisement model from the advertisement information using an image authoring tool, for example, photoshop, and then removes the face image of the user. Insert into the ad model.

Therefore, the conventional method as described above is a two-dimensional transformation technology that replaces the model face in the advertisement information, which is a planar image, with the user's face, which is a planar image, and thus cannot satisfy the desires of users who wish to be more spatial and dynamic. have.

In addition, the conventional method has a problem that it is difficult to replace the face of use when the rotational transformation of the user's face image and the advertisement information is not similar.

Accordingly, an object of the present invention is to provide an apparatus and method for replacing a 3D video model for converting an advertising model of moving picture advertisement information into a face of a specific person.

1 is a schematic configuration diagram of a 3D video model replacement apparatus according to the present invention.

2 shows a three-dimensional standard avatar mesh.

3 is a diagram schematically illustrating a process of generating a front image from a photographic image.

4 is a view showing a principle of extracting three-dimensional information such as face position and rotation angle in the input picture.

5 is a view schematically showing a model face replacement process by the extracted face parameters.

6 is a view showing the result of skin color adjustment of the replaced three-dimensional personal photo avatar.

7 is a view showing a moving picture frame in which a model face of three-dimensional moving picture information is replaced according to the present invention.

8 is a flowchart for performing a three-dimensional moving image model face replacement according to the present invention.

9 is a flowchart for performing face information extraction of a moving image model according to the present invention.

FIG. 10 is a flowchart for performing three-dimensional personal photo avatar transformation by facial parameters extracted through the process of FIG. 9. FIG.

<Description of the symbols for the main parts of the drawings>

100: 3D personal photo avatar generator 102: model face region extraction unit

104: model face parameter extraction unit 106: photo avatar control unit

108: face image converter

In order to achieve the above object, the present invention generates a three-dimensional personal photo avatar according to a photo including a specific face in order to replace the model face in the moving image information with a specific face, and extracts a face parameter of the model face from the moving image information. 3D moving picture model face replacement method and apparatus for modifying the 3D personal photo avatar by the face parameter of the extracted model face and replacing the modified 3D personal photo avatar with the model face image area in the moving image information There is this.

In addition, the face parameter of the present invention includes the position, direction, size, facial expression change, skin color and lighting environment information of the model face.

In addition, face parameter extraction of the model face according to the present invention separates the moving image information for each frame, separates the model face image and the background / foreground image of each frame, thereby validating the flag of the region where the model face image is separated in each frame After extracting any feature points specified on the separated model face images in three-dimensional space, the position, direction, and size information of the model face image is extracted, and the expression change information is extracted by comparing the model face image with the expressionless face image. In addition, there is another feature in extracting skin color information and lighting environment information of the model face image.

Hereinafter, a preferred embodiment of an apparatus and method for replacing a 3D video model according to the present invention will be described with reference to the accompanying drawings.

1 is a schematic configuration diagram of an apparatus for replacing a 3D video model according to the present invention, FIG. 2 is a diagram illustrating a 3D standard avatar mesh, and FIG. 3 is a diagram illustrating a process of generating a front image from a photographic image.

FIG. 4 is a diagram illustrating a principle of extracting three-dimensional information such as a face position and a rotation angle from an input photograph, FIG. 5 is a diagram schematically illustrating a model face replacement process by the extracted face parameter, and FIG. FIG. 3 shows the skin color control result of the personal photo avatar.

7 is a view showing a moving picture frame in which the model face of the 3D moving picture information is replaced according to the present invention.

As shown in FIG. 1, the apparatus for replacing a 3D video model according to the present invention includes a 3D personal photo avatar generator 100, a model face region extractor 102, and a model face parameter extractor 104. And a face image converter 108.

The 3D personal photo avatar generating unit 100 is configured to extract geometric information of a face from a photographic image input from a certain user, and generate a 3D personal photo avatar by modifying a 3D standard avatar mesh according to the extracted geometric information. do.

Here, the input photographic image is a front-like orientation photographic image.

The model face region extraction unit 102 is configured to separate the background or foreground and the face of the person, ie, the face of the model, in each frame of the moving image information by using a predetermined image tool.

In addition, the model face region extractor 102 is configured to separately store the face image and the background / foreground image of the model separated from each frame, and to validate the region from which the model face image is extracted.

The model face parameter extractor 104 is configured to extract face region information such as the position, direction, size, facial expression, skin color, and lighting environment of the separated model face image.

The photo avatar controller 106 is configured to transform the 3D personal photo avatar generated by the 3D personal photo avatar generator 100 by the face parameter extracted from the model face parameter extractor 104.

In addition, the face image converter 108 is configured to map the modified three-dimensional personal photo avatar from the photo avatar controller 106 to an area in which a flag of each frame of the moving picture information is valid to generate a specific converted moving picture.

The operation of the 3D video model replacement device according to the present invention configured as described above is as follows.

First, the 3D personal photo avatar generator 100 generates a 3D personal photo avatar based on photo information input from a predetermined user, which will be described in detail.

That is, the 3D personal photo avatar generator 100 receives front photo information and side photo information as shown in FIG. 3A from a user. The input photo information is converted into a digital image by a scanner or a digital camera. Digital picture information taken by the user.

In addition, as shown in FIG. 3B, the 3D personal photo avatar generator 100 receives a plurality of reference line information for mapping the 3D standard avatar mesh as shown in FIG. 2 from a user on the input digital image.

In this case, the photo input from a specific user generally does not photograph the face front intentionally while taking a photo, and even if the face is intentionally photographed, an accurate front face cannot be photographed.

In addition, a baseline that passes six feature points (two outer edges of both eyes, two end points of the lip, a nose midpoint, and a center top of the lip) on the input picture to map the three-dimensional standard avatar mesh to the input front-facing picture. Receive the specified information from the user.

In this case, the reference line designation information is information designated by the user, a first reference line passing through two points on the outer edges of both eyes, a second reference line passing through two points on both lips, and a nose weight point indicating the direction of the nose and the lips And third baseline information passing through the central dot of the.

Then, six feature points specified in the picture and predefined predefined points on the three-dimensional standard avatar mesh (two outer edges of both eyes on the three-dimensional standard avatar mesh, both ends of the lip, the middle point of the nose, and the upper point of the center of the lips) Mapping to extract geometric information such as camera position, rotation and focal length.

This will be described in detail as follows.

First, the mapping of the 2D photographic image to the 3D standard avatar mesh assumes the condition that the user photographed photographic image is generated when the 3D standard avatar mesh is projected according to the geometrical conditions of the cameras photographing the front and side similar directions. .

Subsequently, under the above assumptions, the projection relationship between the 3D standard avatar mesh and the input photographic image may be defined as a mathematical transformation model (Perspective Transformation matrix) generally used in 3D computer graphics.

Where s is the size (zoom), f is the focal length of the camera, , , Is the three-dimensional rotation angle, , , Represents the moving distance of the camera.

2D points on the photo you specify ( , ) And geometrical information (s, f, , , , , , ).

At this time, by using a nonlinear minimization algorithm, the distance difference between the feature point on the 3D standard avatar mesh and the user-specified feature point on the picture is repeatedly minimized to extract 3D geometric information such as the 3D position, direction, and size of the camera.

Here, in the state where the avatar is fixed by the facts generally known in 3D graphics, the geometric change of the camera is the same as the geometric information of the moving avatar while the camera is fixed, and thus the avatar is extracted by the geometric information of the extracted camera. Extract geometric information of.

The accuracy of the position and orientation of the 3D personal avatar is determined by the accuracy of the feature points specified by the user, and until the user visually judges the geometry of the person in the picture and the geometrical appearance of the 3D standard mesh, the above process is accurately mapped. Is repeated.

Here, the 3D standard avatar mesh as shown in FIG. 2 is configured using a 3D scanner or a professional 3D modeling tool, and since it is already known, a detailed description thereof will be omitted.

In addition, the user obtains an approximation model without having to specify each region in detail, and extracts the exact position of each region through an outline extraction algorithm that uses a change in color values around feature regions that are well known in image processing technology.

In this case, the contour extraction algorithm is an algorithm that extracts the contour by recognizing the portion as a boundary when the color of the pixel suddenly changes beyond the threshold by searching eight peripheral pixels for one pixel in the input image information.

In this case, the threshold value can be arbitrarily set by the user. If the threshold value is small, a large number of boundary lines can be extracted, but unnecessary noise data may be included. If the threshold value is large, only a small number of boundary lines are extracted. It can't be extracted.

In addition, the 3D personal photo avatar generator 100 deforms the 3D standard avatar mesh according to the face position and rotation angle of the feature point of the extracted input image information as shown in FIG. Extract geometric information of.

Next, the 3D personal photo avatar generator 100 controls the 3D standard avatar mesh based on the extracted geometric information to generate a temporary avatar (FIG. 3D).

In addition, the generated temporary avatar is rotated and moved according to the position information and the rotation angle information of the extracted image information to be positioned in the forward direction.

Thereafter, the 3D personal photo avatar generator 100 generates a texture based on the face image of the input photo and generates a 3D personal photo avatar by mapping the generated texture to a temporary avatar (FIG. 3E).

Meanwhile, the 3D personal photo avatar generation unit 100 may generate a 3D personal photo avatar using only the front picture as shown in FIG. 3A. The area corresponding to the side face which originally used the side picture as a texture may also process the front picture. It is made possible by using

A model face extraction operation from moving picture information according to the present invention performed simultaneously with the generation of the three-dimensional personal photo avatar through the above operation will be described.

First, the model face region extraction unit 102 separates moving picture information into individual frames by using a moving picture editing tool. In general, all moving image information includes frame information and frame number information (FPS) displayed on a screen per second, and frames are separated as the frame information is extracted. Here, the reason for separating the frames is to separate the moving image information constituted by a huge amount of still images (30 still images per second) into respective still images. In this case, the frame separation of the moving picture information may be performed by a predetermined professional moving picture editing tool such as Adobe Premier.

Subsequently, the model face region extractor 102 uses the editing tool described above to perform original matching of each frame through template matching based on features such as skin color, eyes, nose, mouth or eyebrows of a human face. After separating the background / foreground and the moving image model from the video information, the background / foreground image and the model face image are respectively stored.

Here, the background of the moving picture information is a background excluding a moving image model, and the foreground means that the face of the moving image model is hidden, for example, a model's hand, a hat, a bottle, and other props such as a mobile phone.

At this time, in the area where the face image is separated from the original frame, the area flag according to the setting of the color of a certain value, such as white, is validated so that the face area to be replaced in the moving image replacement process can be automatically detected in the program. .

Then, the model face parameter extractor 104 extracts face region information of the model face image separated from the moving image information.

That is, as shown in FIG. 4, the model face parameter extracting unit 104 separates the positions of feature points, for example, eyes, nose, and mouth, which are randomly designated on the model face image stored separately from the moving image information. Reverse-projection from the plane into three-dimensional space.

In addition, the model face parameter extracting unit 104 minimizes the distance difference between the feature point on the model face image and the feature point of the three-dimensional standard avatar mesh by a nonlinear minimization algorithm, while the three-dimensional position, direction, and size of the moving image model face image. Extract the parameters.

Subsequently, the model face parameter extractor 104 compares the face of the expressionless state with the model face image of each frame and extracts a facial expression change parameter according to how much in any direction the specified feature points are deformed.

That is, the model face parameter extracting unit 104 may change the facial expression change parameter according to the deformation vector, which is a difference value based on the 'how much the tail is raised' or 'how close the eyes' of the model face image with respect to the face in the expressionless state. Extract

Here, the expressionless face is a face in which predetermined feature points have expressionless face state information and is a face having a predetermined value by a predetermined image editing tool.

The model face parameter extracting unit 104 extracts the lighting environment and the skin color of the model face area, and models the face such as the location, direction, size, lip movement and deformation vector of the extracted model face, lighting environment and skin color information, and the like. The parameter is output to the photo avatar controller 106.

In addition, the model face parameter extracting unit 104 stores the extracted model face parameter in a corresponding script of each frame.

The photo avatar controller 106 controls the position, the viewing direction, the size, the facial expression, the skin color, and the lighting environment of the 3D personal photo avatar by the model face parameter extracted by the model face parameter extractor 104. It demonstrates with reference to.

That is, in order to replace the face of the advertisement model in the advertisement information as shown in FIG. 5A, the 3D personal photo avatar generated as shown in FIG. 5B is placed on the white background image, and the model face extracted by the model face parameter extractor 104 is located. The location information of the 3D personal photo avatar is adjusted to the replacement area.

Then, the direction and the size of the three-dimensional personal photo avatar are adjusted by the direction and the size of the extracted model face.

In addition, the photo avatar controller 106 adjusts the movements of the eyes, eyebrows, and lips based on facial expression information among the model face parameters extracted by the model face parameter extracting unit 104 and stored in the script of each frame to model the moving image information. The expression of the 3D personal photo avatar is adjusted to have the same facial expression.

Also, the photo avatar controller 106 extracts a correction value of the skin color by applying the skin color information of the 3D personal photo avatar to Equation 2 based on the skin color information among the model face parameters.

B = (101-F) / 100

Here, F is a blending parameter, B is a blending vector value. If F is greater than 100, F is set to 100, and if F is less than 1, F is set to 1.

V is the corrected skin color value, A is the skin color value of the 3D personal photo avatar, D is the skin color value of the model face, and subscripts r, g, and b represent RGB channel values for each color.

Also, in general, RGB has a value of 0 to 255 on a computer, where 255 of Equation 1 is the maximum value of RGB.

The photo avatar controller 106 adjusts the skin color of the three-dimensional personal photo avatar based on the skin color value corrected by Equation 2, and simulates virtually the extracted lighting environment information to simulate the model face of the moving image information. Create a 3D personal photo avatar.

On the other hand, when the photo avatar controller 106 replaces the 3D personal photo avatar without considering the lighting environment in one frame of the moving image as shown in FIG. 6A, the boundary line of the face region is clearly displayed as shown in FIG. 6B.

Considering the lighting environment, as shown in FIG. 6C, the replaced three-dimensional personal photo avatar is naturally matched to the surrounding environment.

The face image converter 108 scans each frame of the moving image information to find a portion where the replacement area flag is enabled, and replaces the three-dimensional personal photo avatar output by the photo avatar controller 106 with the new image frame. (FIG. 5C).

Here, the operation of one still image frame has been described, but it is obvious that other frames for moving picture composition perform the same operation.

Through the above operation, each frame in which the model face is transformed into a 3D personal photo avatar in the moving image information is combined to generate a specific converted moving image based on the moving image format (eg, MPEG or AV1).

That is, according to the 3D moving image model replacement apparatus according to the present invention, when moving the model face to the user's face in the moving image information having the frame as shown in FIG. 7A, the moving image information of the frame as shown in FIG. 7B is generated.

A method for performing 3D model registration according to the 3D model replacement apparatus according to the present invention configured and operated as described above will be described with reference to the accompanying drawings.

8 is a flowchart for performing a three-dimensional moving image model face replacement according to the present invention.

First, photo information is input from a predetermined user, and a 3D personal photo avatar is generated according to the input photo information (S800).

That is, the front and side similar direction picture information of the user for generating the avatar is input, and the position and rotation angle information of the face are extracted from the input picture information to match the 3D standard avatar mesh with the face direction of the input picture information. Extract the corrected frontal image as it rotates and moves.

On the other hand, it is possible to create a three-dimensional personal photo avatar that guarantees a certain quality only by the front picture. At this time, the area corresponding to the side face that used the original side photo as a texture is also generated by using the front photo as a texture.

Subsequently, the personal avatar mesh is generated by modifying the 3D standard avatar mesh by modifying the position on the xy plane of the predetermined facial feature point on the corrected front image and modifying the position on the yz plane of the predetermined feature point in the side image.

After matching the common part of the corrected frontal image, the matched image is joined to generate a panorama texture, and the generated personal avatar mesh and the panorama texture are integrated to generate a 3D personal photo avatar.

In addition, the model face image and the background / foreground image are separately stored from the moving image information selected by the user (S802).

Subsequently, a model face parameter is extracted through circular comparison of the model face images (S804), and the 3D personal photo avatar is modified by the extracted model face parameters (S806).

The model face is replaced with a specific face as the modified 3D personal photo avatar is mapped to the model face image area in the moving image information at step S806 (S808).

The model face parameter extraction process of the three-dimensional model replacement process according to the present invention as described above will be described in more detail with reference to FIG. 9.

9 is a flowchart for performing face parameter extraction of a moving image model according to the present invention.

First, frame information in the moving picture information selected by the user is separated for each frame (S900).

Subsequently, the background / foreground image of the moving image information and the model face image are separated from each frame using the image tool (S902).

That is, assuming that the subject's face must exist in each frame, the model face is extracted by extracting an area that is considered similar to the human face on each frame by using a circular comparison based on the features of the human face. Separate.

In this case, the region flag is validated to distinguish it from other regions by setting the region where the model face image is extracted from the moving image frame, that is, the moving image replacement region to a color having a constant value.

From the model face image separated in the above step S902, position, direction, and size information of the model face in the frame is extracted by performing reverse projection of predetermined predetermined feature points on the three-dimensional space from the two-dimensional image plane (S904). .

Subsequently, the facial expression information according to the degree of change in the shape of the mouth and the eye is extracted by comparing the separated face of the model with a preset expressionless face (S906).

In addition, the skin color and lighting information of the model face is extracted (S908).

A three-dimensional personal photo avatar transformation process by the extracted model face parameter during the three-dimensional model registration process according to the present invention will be described with reference to the accompanying drawings.

FIG. 10 is a flowchart for performing 3D personal photo avatar transformation using the model face parameter extracted through the process of FIG. 9.

First, the generated 3D personal photo avatar is placed on a white background image having the same resolution as the moving picture frame, and the position, direction and size of the 3D personal photo avatar are determined by the extracted position, direction and size information of the model face. The back is deformed (S1000).

In operation S1000, the expression of the 3D personal photo avatar is modified by adjusting the shape of the lips and eyes by using the extracted expression information of the 3D personal photo avatar, in operation S1002.

Subsequently, the 3D personal photo is extracted by applying the skin color of the 3D personal photo avatar whose position, direction, size and expression are transformed like the model face of the moving picture frame and the skin color of the extracted model face to a predetermined equation. The skin color of the avatar is adjusted (S1004).

In addition, the lighting environment of the 3D personal photo avatar is controlled by virtually simulating the extracted lighting information (S1006).

An apparatus and method for replacing a 3D moving image model according to the present invention generates a 3D personal photo avatar of a user from photo information, extracts a model face parameter from a model face image in the moving image information, and extracts a 3D personal photo avatar. It transforms by the parameter and replaces it with the model face image area in the moving image information.

Therefore, the present invention has the effect of increasing the advertising effect due to the reuse of the moving picture advertisement by replacing a certain user face in the moving picture information.

In addition, there is an effect that the user can satisfy the desire to participate in the CF or movie by replacing the characters of the moving picture, such as CF or movie with the face of the user.

Although the above has been described with reference to a preferred embodiment of the present invention, those skilled in the art to which the present invention pertains without departing from the spirit and scope of the present invention as set forth in the claims below It will be appreciated that modifications and variations can be made.

Claims (13)

A three-dimensional moving image face replacement method for replacing a first face in a moving image information with a second face, Generating a 3D personal photo avatar corresponding to the photo including the second face; Extracting face parameters of the first face from the moving image information; Deforming the 3D personal photo avatar by face parameters of the extracted first face; Replacing the modified three-dimensional personal photo avatar with a first face image region in the moving image information 3D video model replacement method comprising a. The method of claim 1, And the face parameter comprises at least one of position, direction, size, facial expression change, skin color, and lighting environment information of the first face. The method according to claim 1 or 2, Extracting a face parameter of the first face from the moving image information Separating the moving picture information for each frame; Separating the first face image and the background / foreground image of each frame; Validating a flag of an area in which the first face image is separated in each frame; Extracting position, direction, and size information of the first face image according to back projection of a predetermined feature point designated on the separated first face image into a three-dimensional space; Extracting facial expression change information by comparing the first facial image with the expressionless facial image; Extracting skin color information and lighting environment information of the first face image; 3D moving image model face replacement method comprising a. The method of claim 3, Replacing the modified three-dimensional personal photo avatar with a first face image region in the moving image information And the modified three-dimensional personal photo avatar replaces a frame of each frame of the moving image information with a valid region. The method according to claim 1 or 2, Deforming the 3D personal photo avatar by face parameters of the extracted first face Adjusting the position, direction, and size of the 3D personal photo avatar based on the position, direction, and size information of the first face image; Adjusting an expression of the 3D personal photo avatar by the extracted expression change information; Adjusting the skin color of the 3D personal photo avatar by extracting the skin color of the extracted first face image and the corrected skin color according to the skin color of the 3D personal photo avatar; Adjusting the lighting environment of the 3D personal photo avatar by virtually simulating the lighting environment of each frame based on the extracted lighting environment information. 3D video model replacement method comprising a. The method of claim 5, And the corrected skin color is calculated using the following equation. B = (101-F) / 100 However, F is a blending parameter, where F is 100 if F> 100 and F is 1 if F <1. B is the blending vector value, V is the corrected skin color value, A is the skin color value of the three-dimensional personal photo avatar, D is the skin color value of the first face, Subscripts r, g, b are RGB channel values for each color 255 is the maximum value of RGB. A three-dimensional moving image model face replacement apparatus for replacing a first face in a moving image information with a second face, A 3D personal avatar generation unit generating a 3D personal photo avatar corresponding to the photo including the second face; A face parameter extracting unit extracting face parameters of the first face from the moving image information; A photo avatar controller configured to modify the 3D personal photo avatar by face parameters of the first face; A face image converter which replaces the modified three-dimensional personal photo avatar with a first face image area in the moving image information. Is configured to include a three-dimensional moving image model face replacement device. The method of claim 7, wherein And the face parameter includes at least one of position, direction, size, facial expression change, skin color, and lighting environment information of the first face. The method according to claim 7 or 8, The face parameter extracting unit Separating the moving picture information for each frame; Separating a first face image and a background / foreground image of each frame; Validating a flag of a region in which the first face image is separated in each frame; Extracting position, direction, and size information of a model face image according to back projection of any feature point designated on the separated first face image into a three-dimensional space; Extracting facial expression change information by comparing the first facial image with the expressionless facial image; And a skin color information and an illumination environment information of the first face image. The method according to claim 7 or 8, The photo avatar controller Adjust the position, direction, and size of the 3D personal photo avatar based on the position, direction, and size information of the first face image; Adjusting an expression of the 3D personal photo avatar by the extracted expression change information; Adjusting the skin color of the 3D personal photo avatar by extracting the skin color of the extracted first face image and the corrected skin color according to the skin color of the 3D personal photo avatar; And virtually simulating the lighting environment of each frame based on the extracted lighting environment information to adjust the lighting environment of the 3D personal photo avatar. The method of claim 10, And the corrected skin color is calculated using the following equation. B = (101-F) / 100 However, F is a blending parameter, where F is 100 if F> 100 and F is 1 if F <1. B is the blending vector value, V is the corrected skin color value, A is the skin color value of the three-dimensional personal photo avatar, D is the skin color value of the first face, The subscripts r, g, and b represent the RGB channel values for each color, 255 is the maximum value of RGB. A three-dimensional moving picture model face replacement method for replacing a first face in a moving picture information with a second face, Generating a 3D personal photo avatar according to the photo including the second face; Extracting a face parameter of the first face from the moving image information Separating the moving picture information for each frame; Separating the first face image and the background / foreground image of each frame; Validating a flag of an area in which the first face image is separated in each frame; Extracting position, direction, and size information of the first face image according to back projection of a predetermined feature point designated on the separated first face image into a three-dimensional space; Extracting facial expression change information by comparing the first facial image with the expressionless facial image; Extracting skin color information and lighting environment information of the first face image; Adjusting the position, direction, and size of the 3D personal photo avatar based on the position, direction, and size information of the first face image; Adjusting an expression of the 3D personal photo avatar by the extracted expression change information; Adjusting the skin color of the 3D personal photo avatar by extracting the skin color of the extracted first face image and the corrected skin color according to the skin color of the 3D personal photo avatar; Adjusting the lighting environment of the 3D personal photo avatar by virtually simulating the lighting environment of each frame based on the extracted lighting environment information; Replacing the modified three-dimensional personal photo avatar with a first face image region in the moving image information 3D video model replacement method comprising a. A recording medium in which a program of instructions executable by a digital processing apparatus is tangibly implemented to perform a three-dimensional moving image model face replacement method, and which can be read by a digital processing apparatus, The 3D video model replacement method, Generating a three-dimensional personal photo avatar according to a photo including a specific face; Extracting face parameters of a model face in the moving image information; Deforming the 3D personal photo avatar by face parameters of the extracted model face; Replacing the modified three-dimensional personal photo avatar with a model face image area in the moving image information Recording medium comprising a.