KR20010091743A - A formation method of an automatic caricature - Google Patents

Abstract

PURPOSE: A method is provided to generate caricature for facial features in an easy and automated manner. CONSTITUTION: A method comprises a step(S10,S20) of inputting user's face, converting the input user's face into a color model and removing unnecessary noise; a step(S30) of detecting areas for each part of the user's face; a step(S40) of extracting personal facial features by comparing a prediction algorithm and data stored in a database with respect to the detected areas of each part of the user's face; a step(S50) of personifying by transforming and combining the extracted personal facial features; and a step(S60-S80) of generating caricature if each element of the areas of personified face is judged as appropriate in view of time and space, and re-combining each element and generating caricature if each element is judged as inappropriate.

Description

{A formation method of an automatic caricature}

The present invention relates to a method for generating a caricature, and more particularly, an automatic caricature generation method for generating a unique caricature by extracting a face region and a feature from an input image and comparing the feature points according to each part with a pre-stored database. It is about.

Characters are characters of characters, personalities and personalities of characters, characters, characters, symbols, etc., and are characters in dramas, movies, or characters.

Characters are divided into two types, which are classified into a dictionary meaning character and a marketing meaning character.

Characters with a characteristic of a substance are called characters in a dictionary meaning. Characters in a marketing meaning are expressions of phenomena having vitality created for the purpose of use under artificial and intentional strategy to create high added value through pictures or photos. It can be said.

The strength of the character industry is, firstly, the sale of non-age, non-sex, and non-generational, intangible elements that can instrument consumer appeal, not just sales of characters alone. .

Second, it is possible to differentiate brands by maximizing and diversifying items and to reorganize the overall shop in a total-shop rather than a single brand. The Japanese Disney Store, for example, comprises an average of 3,000 products.

Third, it is not only limited to comprehensive entertainment products, but also expands entertainment facilities such as various media, entertainment, and resorts, and plays an important role in creating synergy effects in character business. For example, you can choose Disney cartoons and Disneyland.

In addition, the character enhances the attractiveness of the character as the subject of the advertisement because it has the characteristic of forming a wide consensus among all people beyond generation and nationality regardless of gender and age.

In general, most commercials use celebrities as models. However, no matter how popular the actor, there are people who do not like it, and as you age, you naturally lose your charm. By comparison, characters don't lose their charm over time, but rather can be redeveloped, reflecting social change. For example, Disney's characters are still loved today, decades after their development, and are still being reborn in a new way to suit reality. In addition, despite being developed in the United States, it is one of the most beloved characters in the world, including our country.

Therefore, the use of characters is essential for new start-ups, newly launched products, and many competitors.

However, characters are accompanied by some pictorial skills. Until now, people have drawn each one by hand. Like this, the characters drawn by pens or paintbrushes have the advantages of drawing with artistry, sophistication, and unique characteristics, but they have limitations in applicability and expandability in places, times, and available people. In addition, even in view of the current commercial character, due to the manual work by order, there is a lot of trouble, such as the cost and time required.

Therefore, the present invention was devised to solve the above-mentioned problems, and an object of the present invention is to extract feature points by extracting facial regions from an image of a person input using a scanner or a video camera, and to perform deformation operations for each element. The present invention provides a generation method for automatically generating an anthropomorphic caricature.

According to an aspect of the present invention, there is provided a system in which a user inputs an image to generate a caricature, the preprocessing operation step of converting a user's face into a color model and removing unnecessary noise, and the preprocessing operation. Detecting an area of each part of the face after performing the step, and extracting a feature point by comparing and analyzing the area of each part of the detected face with a prediction algorithm and information previously stored in a database; A personification step of modifying and combining the extracted feature points and anthrottizing each element of the personified face region in terms of time and space to determine suitability, and if not suitable, generates a caricature and recombines each element to generate a caricature. Characterized in that the caricature generation step.

1 is a system block diagram for performing the automatic caricature generation method of the present invention,

2 is a flowchart of an automatic caricature generation method according to the present invention;

3A is a flowchart of a genetic algorithm for detecting a face region of FIG.

3B is a flowchart for extracting the feature point of FIG. 2;

3C is a flowchart of the personification operation of FIG. 2;

FIG. 3D is a flowchart of correction of the generated caricature of FIG.

<Description of Symbols for Main Parts of Drawings>

10: input device 20: database

30: control unit 40: caricature generation unit

50: output device

Hereinafter, an automatic caricature generation method of the present invention will be described in detail with reference to the accompanying drawings.

1 is a system block diagram for performing the automatic caricature generation method of the present invention.

Referring to FIG. 1, an input device 10 for inputting a face image to a scanner or various cameras, a database 20 previously storing feature points for each part of a face, and a face region in the input image. A caricature generation unit 40 which extracts and compares the data of the database 20 to generate a caricature, an output device 50 which outputs the result of the caricature generation unit 40, and the caricature generation unit 40 And a controller 30 for controlling the database 20 and the entire input / output device.

2 is a flowchart of an automatic caricature generation method according to the present invention, FIG. 3A is a flowchart of a genetic algorithm for detecting a face region, FIG. 3B is a flowchart for extracting feature points, and FIG. 3C is a flowchart of anthropomorphism. 3d is a correction flowchart of the generated caricature.

2 and 3a, 3b, 3c, and 3d will be described in detail the operation of the automatic caricature generation method of the present invention.

First, a face of a user who wants to caricature is input to the input device 10, that is, a scanner or various cameras (S10).

The input image is converted into a color model and a preprocessing operation for removing unnecessary noise is performed (S20). Then, after performing the pretreatment operation, an area for each part of the face is detected (S30).

Detection of the face region from the image is a process of dividing only the face from the image, and the present invention uses a genetic algorithm.

Evolutionary algorithms are computational models that attempt to solve complex real-world problems by simulating the evolutionary processes of the natural world on a computer. Evolutionary algorithms have various methods depending on the method of representing chromosomes and the types and characteristics of operators. You can share it.

First, Genetic Algorithms (GA) use fixed-length binary strings as chromosomes, and Evolution Strategies (ES) use vectors consisting of genes that take real values. do. In addition, there are Evolutionary Programming (hereinafter referred to as EP) and Genetic Programming (hereinafter referred to as GP) that use graphs and trees to express dyes. The operators used in evolutionary search are mainly EP and ES for mutations, and GA and GP for crossovers.

In the evolution of living things in the natural world, there is a high probability that a collection of individuals forming a generation, that is, a population with high fitness for the environment, survives and reproduces with a high probability. And crossover and mutation, forming the next generation of populations.

GA is an algorithm that artificially models the evolution of the above organisms.

Due to the features described above, it is more likely to obtain a global solution than a calculation-based method (hill-climbing), which is one of other search or optimization methods, and is more efficient than other search methods.

Therefore, in the present invention, by generating a portion of the face as an initial group (S31), it is evaluated whether appropriate (S32). In addition, a selection, evaluation (S33) is performed according to the evaluated result, and a cross mutation process of binding to other stored data is performed (S34). After the cross mutation process, the fitness is again evaluated (S35) to determine whether it satisfies a predetermined criterion to extract a region for the portion (S36). Of course, if the above is not satisfied, go back to step S32 to select another data and evaluate again.

If an area for each part of the face is detected as a result satisfied in step S36, feature vector extraction is required from this (S40).

When the face region, that is, the outline information of the eyes, the eyebrows, the lips, and the face is extracted (S41), the extraction information through the prediction algorithm is analyzed (S42), and compared with data for each face region previously stored in the database 20. Analyze the extraction information through (S43).

According to the results of the comparative analysis (S42, S43), the user's personal feature points are extracted (S44).

Then, the personification operation (S50) of deforming and combining the extracted feature points, that is, each element is performed.

In the personification operation S50, as shown in FIG. 3C, the priority of the extracted personal characteristics is determined (S51), and the emphasis (incidence) rate for each feature point is determined according to the priority of each element determined in terms of time and space. Determine (S52).

After deforming and applying each element to each element according to the determined emphasis (incidence) ratio (S53), the modified elements are integrated (S54) to generate a carrier (S60).

In the generated caricature, a caricature is completed. The user determines the suitability of each element in terms of space and time as shown in FIG.

If each element is not appropriate, the operation of correcting each element for changing the position of each element is performed (S72) and recombined again (S73).

When the user satisfies the caricature generated by the recombination is output through the output device 50 (S80).

As described above, according to the present invention, even a non-expert can simply generate a caricature of a human face to perform a simplification required for the production of a character, pictorial pictures of simple sketches, and animation, and the result of the work It can be immediately confirmed, which has the effect of activating the application of commercial advertising and the caricature industry.

Claims (2)

In the system for generating a caricature by inputting an image, A preprocessing operation step of converting a user's face into a color model and removing unnecessary noise; Detecting an area of each part of the face after performing the pretreatment operation; A feature point extraction step of extracting a personal feature by comparing and analyzing a prediction algorithm and information previously stored in a database for each area of the detected face; A personification step of modifying and combining the extracted feature points and personifying them; And a caricature generation step of generating a caricature if suitable by determining the suitability of each element of the personified face region from a space-time point of view, and recombining the respective elements if not suitable. The method of claim 1, wherein the personification step, Determining a priority of the extracted personal characteristics; Determining the emphasis (incidence) rate of the feature according to the determined priority; Applying and modifying each element according to the emphasis ratio; Automatic caricature generation method comprising the step of integrating each of the modified elements.