KR20030042403A - Facial character manufacturing method by fitting facial edgeline - Google Patents

Abstract

PURPOSE: A method for generating a face character by contour matching is provided to easily form a face character by selecting a face photograph image and matching a contour based on the face photograph image. CONSTITUTION: A user inputs a face photograph image(S1). Positions of eyes of the face photograph image are grasped and the face photograph image is normalized based on the positions(S2). A spline curve is approximated to a face contour on the normalized face photograph image and colors of face components are decided(S3). A texture of the face photograph image and a character image are convoluted(S4). The spline curve is expressed in the contour based on characteristic points and the colors, and character face components are generated through anti-aliasing(S5). A face character image is generated and displayed(S6). A complete character image is generated by combining the face character with a hair, clothes, and accessories selected from a sample database of a server(S7). The generated character is stored or transmitted to the server(S8).

Description

Facial character generation method using contour matching {Facial character manufacturing method by fitting facial edgeline}

The present invention focuses on simplifying the process of making a face character, by moving a control point to match the contour of a component on a face picture, approximating a spline curve, and convolution of the texture on the face picture with the skin color of the character face. To create a face character that looks like The present invention has the advantage that it is possible to generate a character of various forms only by changing the spline feature point, and to reflect the texture of the skin on the photo by adjusting the convolution mask. In addition, anti-aliasing was performed to reduce the effect of aliasing that occurs during the expression of characters.

The present invention can be used for the facial expression change or character animation of the character, in particular can be utilized as part of the wireless avatar animation system by face recognition.

Currently, the character generation method is a method in which a professional designer sketches an individual's face to make a character, and extracts a part element of a face photograph and compares it with samples of each part in the database to select and combine similar samples to create a character. There is a way. In the former case, a lot of effort and time are required, and the character is generated by the designer. In the latter case, many sample images should be prepared, and the human face is the body part where the personality is most strong, and there are limitations in expressing various faces no matter how many samples.

The present invention provides a method of generating a face character image resembling a photographic image only by adjusting a control point of a spline curve without the help of a professional character designer or a sample image database for combining images of eyes, nose, mouth, etc. for generating a face character. do.

A first step of receiving a face photograph image by a user; A second step of grasping both pupil positions of the face photograph image by a manual or automatic method and normalizing the face image (size and rotation angle) based on the position; A third step of approximating a spline curve over the facial contour on the normalized picture and determining the color of the facial components; Based on the fourth and fourth stages of convolution of the texture and character image of the face photograph image, the spline curve is outlined based on the feature points and colors obtained in the third stage, and the components of the character's face are expressed through anti-aliasing. A fifth step; A sixth step of finally generating a face character image and displaying the image; A seventh step of selecting a hair, a costume, and other accessories from a sample database of the server and combining the generated hair character with the generated facial character; And an eighth step of storing or transmitting the character generated by the seventh step to the server.

1 is a block diagram of a character generation system for carrying out the present invention.

2 is a flowchart illustrating a process of generating a character according to the present invention.

Hereinafter, with reference to the accompanying drawings will be described in detail an embodiment of the present invention.

A feature of the present invention is to input a face photo image and normalize it based on size and horizontality, and generate a character by approximating a spline curve to the photo image contour by adjusting a control point. Selection can be made using a table. Instead of working on all the points that make up the curve to change the curve, the spline curve feature changes only the characteristic points, and the convolution of the texture and the character image on the face photograph image. Various characters can be created according to the shape, and desired face characters can be generated with little effort and time investment.

1 is a block diagram of a system for generating a character according to the present invention, an input device 1 such as a mouse, a keyboard and an image input device, an image processing device 2 for processing input information and image information, and processed images. Image output device (3) showing the information on the screen, image storage device (4) for storing the image information, communication device (5) for importing or transmitting images from the server-side database via the network, character character storage It consists of an image database (6), a sample image database (7) for storing hair styles, body shapes and accessory images, and a server (8) for web services.

2 shows a flowchart of a character generation process according to the present invention. As shown in the flowchart, in step S1, the face photograph image selected by the user is input to generate the character image.

In step S2, the face part of the input face picture image is automatically or manually obtained from the face part to be generated as a character, and the face picture image is normalized based on the position information of both pupils. In step S2, the size of the face desired to generate the character is deformed to a constant size, and the position of the pupil is rotated into a face image having a horizontal shape.

In step S3, the spline curve is approximated based on the face image adjusted in step S2 of obtaining the shape and position information of each face component. Adjust the spline control points of each face component to match the contour shape of the eyes, nose, mouth, ear, eyebrows, and facial outline with the spline curve, select whether to support double eyelids and open mouths, and create the character face components Select the color of the eyebrows, lips and face from the color table to specify the color. In addition, the degree of reflection is set so that the color and contrast of the input face photograph image can be reflected in the character face image.

In step S4, a convolution process of the texture of the face photograph image and the character face image is performed with the degree of reflecting the face photograph image set in step S3. This process is based on a convolutional mask, and by adjusting each value within the mask, the composite weight of the texture of the face and the color of the character's skin can be determined.

Reflectance index on point (x, y) With If the brightness value in the mask corresponding to the point (x, y), T is the user's adjustment value, the reflectance index is obtained from (Equation 1). The reason why the reflectivity obtained from the mask image and the reflectivity by the user selection are doubled in Equation 1 below is that the mask has a fixed reflectivity that reflects the morphological characteristics of the face, and the reflectivity by the user selection is a character generation. This is to make it easy to reflect the subjectivity of the user of the city.

(Equation 1)

The reflectivity is calculated by the above (Equation 1), and then the color value of the final character face image is calculated using the following (Equation 2).

(Equation 2)

here, Is the character image value at point (x, y) Is the background value at the point (x, y), indicating the face texture on the photo.

In step S5, the face component is expressed based on the image generated through the convolution process of step S4 and the contour information and the color information of each face component obtained through step S3. Spline polynomials are used to obtain the entire spline's components, and through the linearization and anti-aliasing of these components, the contours of the character's face components are represented, and face components are applied by applying the set face color, eyebrow color, and lip color.

In step S3, to express the contours of the facial components, the spline polynomials are used to obtain the whole splines and connect them in a straight line by the Bresenham algorithm. The Bressenham algorithm can be used to represent circles or other curves, using only integer arithmetic to find and draw coordinates close to the actual straight line.

Anti-Aliasing is applied to straight lines to get higher quality character images and to eliminate straight lines and edges. Anti-aliasing improves the quality of the image by making the outline of the character smoother. Actually, the pixel value is set around the straight line according to the gap between the path of the straight line and the pixel position.

The difference in position along the x- and y-axis directions Say Is represented by (Formula 3). the derivative of the x-axis component Axial component derivative Speaking of When is anti-aliased for the y-direction neighbors of the pixel, In this case, anti-aliasing processing is performed on neighboring pixels in the X direction. Also ego , Anti-aliasing is performed on the point (x, y) and the neighbor point (x, y + 1) in the y direction. If it is, the anti-aliasing process is performed on the point (x, y) and the neighbor point (x, y-1).

Point This is called anti-aliased result on Point Pixel value you want to give to, Point Defining the pixel value given in Equation 4 below shows the anti-aliasing process.

(Equation 3)

: Y value by equation of straight line

: Y value by Bressenham linear equation

(Equation 4)

In the case of

In the case of

Step S6 is a step of finally generating a face character image. The face character image generated through steps S4 and S5 is transformed into a predetermined size and then displayed. If it is determined that the generated face character image needs correction, the character face image may be corrected by returning to step S3 by adjusting the outline or changing the color of the face component to be corrected.

In step S7, the head style, body shape, and other accessories are selected from the sample database on the server side, and then combined with the face character image to complete the entire character image.

In step S8, the completed character image is stored in the storage device or transmitted to a web server through a communication device for use of the character image on the web and stored in the character database.

According to the present invention, when a user wants to make a character similar to himself or herself to use as an avatar in a chat or an e-mail, or to express a face of another person such as an entertainer as a character, the user selects a face photograph image based on the selection Contour matching allows face characters to be created on the fly.

Claims (4)

In the face character generation method by matching the contour, A first step of receiving a face photograph image by a user; A second step of grasping both pupil positions of the face photograph image by a manual or automatic method and normalizing the face image (size and rotation angle) based on the position; A third step of approximating a spline curve over the facial contour on the normalized picture and determining the color of the facial components; The fourth step of convolution of the texture and character image of the face photograph image, and the spline curves are outlined based on the feature points and colors obtained in the third step, and the components of the character face are processed through an anti-aliasing process. A fifth step; Finally, generating a face character image and displaying the image; A seventh step of selecting a hair, a costume, and other accessories from a sample database of the server and combining the generated hair character with the generated facial character; And an eighth step of storing or transmitting the character generated by the seventh step to the server. The method of claim 1, Characterization method characterized in that to approximate the face image contour by adjusting the spline control point based on the face image in the third step. The method of claim 1, In the fourth step, the character generation method comprising convolution of the texture of the face photographic image and the character image. The method of claim 1, Using the spline polynomial, the components of the entire spline curve are obtained by using the spline polynomial with the shape and position information of each facial component obtained through the control point adjustment in the fifth step. The contours of the facial components are linearized and anti-aliased. Character generation method characterized in that