KR102094063B1 - Genuine art object judgment method and system using markov chain - Google Patents

Abstract

The present invention relates to a forgery and forgery differentiation system using a Markov chain and a method for differentiating an art, and the image of the object to be assessed is processed through a color type analysis method and an entropy type analysis method to synthesize it and probabilities of authenticity of the artwork. The present invention relates to a system for differentiating and forgery of forgery of art using a Markov chain with high accuracy.
That is, the present invention is a database unit in which an image file of an original artwork is stored and color code signal information and statistical code information about the entropy of the entire original image are stored, and an image extraction unit that acquires image information of the artwork corresponding to the object of discrimination. And, a color type analysis unit that divides each image of the image extraction unit according to a color difference signal and then applies a Markov chain to each pixel to match the color difference information stored in the database unit, and quantizes the entropy of each image of the image extraction unit. And the statistical encoding, and then compares the statistical code information stored in the database with the Entropy type analysis unit that compares the Markov chain method, and the correlation between the color type analysis unit and the result values of the entropy type analysis unit, and similarity with the original artwork. Difference calculation unit for calculating; And that is characterized.

Description

{Genuine art object judgment method and system using markov chain}

The present invention relates to a forgery and forgery differentiation system using a Markov chain and a method for differentiation, and the image of the object to be assessed is processed through a color type analysis method and an entropy type analysis method. The present invention relates to a forgery and forgery differentiation system using a Markov chain with high accuracy.

In recent years, as people's living standards have risen, interest in art works has increased compared to the past, and its value is so high that the auction market, offline or online, has been activated.

As such, as the trade in art works becomes active, cases of forgery and alteration of art works are increasing. Since it is difficult for the general public to discriminate whether these artworks are forged or altered, the discrimination method by experts and the discrimination method using advanced optical equipment are used for more reliable discrimination.

However, the discrimination method by experts judges whether or not authenticity is based on a subjective eye, so it is difficult to obtain high accuracy, and in recent years, forgery and alteration techniques have been increasingly advanced.

In addition, the forgery discrimination method using various analytical equipment such as high-tech optical equipment has the advantage of fine-grained discrimination than the discrimination method by experts, but there are problems such as a problem of damage to an artwork or a long analysis time. Each equipment has its own advantages and disadvantages, and its reliability is not well proven.

As a prior art for discrimination of art works, in Patent Publication No. 10-2013-0141403 (Artwork Appraisal Method and System Using Characteristics of Art), the art property extracting unit extracts the art object of appraisal using appraisal object art information. ; A matching unit matching the extracted characteristics of the object of emotion with the corresponding characteristics by comparing with the characteristics of the authentic or author of the database; A authenticity determining unit determining whether the emotion-targeted artwork is authentic based on the information on the corresponding characteristics; Disclosed is a technical content that includes.

However, in the prior art, it is difficult to discriminate a wide range of works of art because only limited information such as paper type, texture, and fine matter of canvas is included in extracting the properties of works of art. If made, a difficult problem arises in concealing authenticity. In addition, there is a problem that a part of the artwork to be a sample is damaged in the process of extracting the characteristics of the artwork.

As another prior art, the registered patent No. 10-1759247 (system and method for authenticity of artwork using optical characteristic analysis) irradiates white light onto an artwork and irradiates a line laser onto the artwork, and irradiated from the light irradiation unit Optical property measurement and processing unit that measures reflected light information by white light, intensity of light, and image data that appears when the line laser is scattered and reflected, and reflected light information, reflected light intensity information, and surface shape measured by the optical property measurement and processing unit An optical property DB construction unit that DB stores and stores the optical property information of the corresponding art based on data, and an art authenticity determination unit that determines the authenticity of the object to be judged based on the optical property information stored in the optical property DB construction unit A technique including; has been disclosed.

However, since the prior art relies only on the optical characteristics reflected on the artwork, it is difficult to analyze data using the color information of the artwork intuitively, and there is a problem in that the result may vary depending on the performance change of the light irradiation unit.

Patent Publication No. 10-2013-0141403 Registered Patent No. 10-1759247 Patent Publication No. 10-2009-0032637

The present invention was devised for an objective and accurate comparison of the original artwork and the object to be differentiated, and collects information obtained through entropy type analysis by analyzing color type analysis and statistical code information on entropy that contrasts color difference information of the artwork. The object of the present invention is to provide a forgery and forgery discrimination system using a Markov chain that can significantly increase the accuracy of forgery and forgery discrimination by calculating the similarity with the original artwork.

The discrimination system of the present invention is a database unit in which an image file of an original artwork is stored, color difference signal information, and statistical code information about the entropy of the entire original image, and an image extracting image information of an artwork corresponding to the discrimination target A color type analysis unit that divides each image of the image extraction unit according to a color difference signal and then applies a Markov chain to each pixel to match the color difference information stored in the database unit, and the entropy of each image of the image extraction unit. After the quantization and statistical encoding, the entropy type analysis unit that compares the statistical code information stored in the database unit with the Markov chain method, and analyzes the correlation between the color type analysis unit and the result values of the entropy type analysis unit, and the original artwork. Differential calculation unit for calculating the similarity; It is characterized by being made.

In addition, the color type analysis unit; after dividing the image of the image extraction unit into each image screen, the color difference signal division, screen division for each color difference signal, pixel elements of the color difference signal screen through the sequence of pixel elements, the artwork corresponding to the differentiation object It is characterized by obtaining the color difference signal data of the.

In addition, the entropy type analysis unit; divides the image of the image extraction unit into each image screen, and then undergoes color-difference signal transformation, high-speed discrete cosine transformation, quantization, and statistical encoding processes in order for the entropy of the artwork corresponding to the discrimination target. It is characterized by acquiring statistical code data.

In addition, the discrimination method of the present invention comprises the steps of constructing a database of image code of the original artwork and storing color difference signal information and statistical code information about the entropy of the entire original image in a database. Obtaining image information of an art work, dividing each image of the acquired image information according to a color difference signal, applying a Markov chain for each pixel, and comparing the color type information stored in the database unit with the color type analysis step; The entropy type analysis step of comparing each image of the image information with entropy and statistical encoding, and then comparing the statistical code information of the database with the Markov chain method, and the result of both the color type analysis step and the entropy type analysis step Original by analyzing correlation It characterized by consisting of; calculating a degree of similarity between sulpum.

In addition, the color type analysis step; A screen division step of dividing the acquired image information into a plurality of image types to fit a frame, a color difference signal division step of acquiring color difference information of the divided image, and a screen division step of each color difference signal that is divided according to the divided color difference signals. A pixel elementalizing step of acquiring pixel values of each of the chrominance signal screens, applying a Markov chain to the pixels, fetching chrominance signal information from the database, and chrominance signal information and Markov in the database. It characterized in that it comprises a; step of matching the color difference signal information to which the chain is applied.

In addition, after the step of comparing the color difference signal information; then, differentiating each unit process using a specific function, applying multiple layers to the differentiating step, and an object to be discriminated through an error backpropagation algorithm Compensating the color difference signal information of; characterized in that it further comprises.

In addition, the entropy type analysis step includes: a screen division step of dividing the acquired image information into a plurality of image forms according to a frame, a color difference signal conversion step of converting the RGB form of the segmented image into a YUV form, and the converted image. Is a step of undergoing a fast discrete cosine transform (FDCT) process, quantizing in bit form as the result, statistically encoding the quantized data, and statistical code information for entropy in the database It characterized in that it is composed of; and a step of comparing the statistical code information of the database and the statistical coded data of the object to be discriminated using a Markov chain.

In addition, a step of comparing the statistical code information of the database and the statistical coded data of the object to be discriminated using a Markov chain; and then determining the number of dimensions N and then comparing the statistical coded data in a multidimensional (N-dimensional) array. And, it characterized in that it further comprises a; step of collecting the multi-dimensional contrast results.

The present invention is based on information that can be obtained from an image of a work of art and collects information obtained through entropy type analysis of statistical code information about color type analysis and entropy that contrasts color difference information of the work with each other, and is the original for the object to be differentiated. By calculating the degree of similarity with the work of art, objective and accurate judgment can be made, and it has the effect of significantly improving the forgery and discrimination accuracy of the work of art.

In addition, in the color type analysis method, multiple layers are applied in the process of contrasting the color difference signal with the original, and the color difference signal is continuously corrected through an error back propagation algorithm, thereby improving the analysis reliability through correction learning by deep learning. There is.

1 is a configuration diagram showing a forgery and discrimination system for art using the Markov chain of the present invention.
Figure 2 is a block diagram showing the overall discrimination method using the differential system of the present invention
Figure 3 is a block diagram showing the color type analysis step in the differential method of the present invention
Figure 4 is a block diagram showing the entropy type analysis step in the differential method of the present invention

Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings. And in the description of the present invention, if it is determined that a detailed description of related known functions or configurations may unnecessarily obscure the subject matter of the present invention, the detailed description will be omitted.

The present invention was devised for an objective and accurate comparison of the original artwork and the object to be differentiated, and collects information obtained through entropy type analysis by analyzing color type analysis and statistical code information on entropy that contrasts color difference information of the artwork. The present invention relates to a forgery and forgery differentiation system and a method of differentiation using a Markov chain that can significantly increase the accuracy of forgery and forgery discrimination by calculating the similarity with the original artwork.

First, in the art forgery forgery discrimination system using the Markov chain of the present invention, as shown in FIG. 1, an image file of an original artwork is stored, and a database unit storing color difference signal information and statistical code information about the entire entropy of the original image ( 10), an image extraction unit 20 for acquiring image information of an artwork corresponding to a differentiation object, and each image of the image extraction unit 20 according to a color difference signal, and then applying a Markov chain for each pixel The color type analysis unit 30 that contrasts with the color difference information stored in the database unit 10, and after quantization and statistical encoding for the entropy of each image of the image extraction unit 20, and then statistically stored in the database unit 10 Entropy type analysis unit 40 that compares code information with a Markov chain method, and the color type analysis unit and entropy type analysis It is characterized by consisting of; a differential calculation unit (50) for calculating the similarity with the original artwork by analyzing the correlation with the result value of the unit (40).

In the database unit 10, an image file for an original artwork is stored, and information on the corresponding image, that is, color difference signal information and statistical code information about the entropy of the entire image are stored.

In addition, the image extraction unit 20 is configured to acquire image information of an art object corresponding to a discrimination target through a separate image photographing equipment. The information obtained based on the photographed image is compared with the color difference information of the database unit 10 and statistical code information.

The analysis configuration of the present invention is largely divided into a color type analysis unit 30 and an entropy type analysis unit 40. The color type analysis unit 30 expresses the pixel elements of each of the divided color difference signal screen RGB as a state space function, and the information on the color difference signals therebetween is the color difference signal information of the original artwork existing in the database unit 10. It has a structure that contrasts with. In addition, the entropy type analysis unit 40 is configured to contrast the pattern between the statistical code value of the entropy of the divided spatial coordinate region and the statistical code value of the original artwork existing in the database unit 10.

The color type analysis unit 30 divides the image of the image extraction unit 20 into respective image screens, and then sequentially through the process of pixel element splitting of the color difference signal, screen division by color difference signal, and color difference signal screen. It is characterized by acquiring color difference signal data of an artwork corresponding to a discrimination object. The chrominance signal division, the screen division for each chrominance signal, and the pixel elementalization process of the chrominance signal screen are a series of processes for obtaining element information of a pixel that is the basis of the chrominance signal data, and the chrominance signal data thus obtained is a Markov chain for each pixel. By applying, it is possible to contrast with the color difference signal information of the database unit 10.

Also, the entropy type analysis unit 40 divides the image of the image extraction unit 20 into respective image screens, and then performs chrominance signal transformation, fast discrete cosine transformation, quantization, and statistical encoding processes in order to differentiate objects. It is characterized by acquiring statistical code data for the entropy of the corresponding artwork. The chrominance signal conversion, the fast discrete cosine transform, and the quantization process are a series of processes for obtaining statistically encoded data about the entropy of an image, and this process is similarly used in the field of image compression technology.

The Markov Chain used in the present invention means a discrete-time stochastic process that was introduced by Russian mathematician Andrei Markov in 1906 and has Markov Property. Here, the Markov property means that the conditional probability distribution of the future state when the past and present state is given is determined only by the present state independent of the past state. That is, the state transition probability matrix is the key, and the probability distribution can be calculated for the number of each case using the probability. The probability distribution using this Markov chain has the characteristic that the accuracy increases as the number of sampling increases.

The differential calculation unit 50; analyzes the correlation between the color type analysis unit and the result value of the entropy type analysis unit 40 to calculate similarity with the original artwork. Through the similarity calculated in this way, it is finally judged whether or not the object of discrimination is forged.

Below, the discrimination method using this system of the present invention will be described in more detail.

The discrimination method of the present invention comprises the steps of building an image file of an original artwork as shown in FIG. 2 and constructing color difference signal information for the corresponding image and statistical code information for the entropy of the entire original image in a database (S-1). And, step (S-2) of acquiring the image information of the artwork corresponding to the object to be discriminated through the image equipment, and dividing each image of the acquired image information according to a color difference signal, and applying a Markov chain to each pixel to apply a database The color type analysis step (S-3) of contrasting the color difference information stored in the unit, and each image of the acquired image information is quantized and statistically encoded for entropy, and then entropy that is compared with the statistical code information in the database using the Markov chain method. Both the type analysis step (S-4) and the color type analysis step (S-3) and entropy type analysis step (S-4) Comprising the step of calculating the similarity with the original artwork by analyzing the correlation to the excess value (S-5);

The step (S-1) of constructing in the database is a process of storing not only the image files of each original artwork, but also statistical information about color difference signals and entropy of the corresponding image files, and obtaining the image information. Step (S-2); is to acquire the image information of the art object corresponding to the discrimination object through a separate image recording equipment.

The color type analysis step, as shown in Figure 3, the color type analysis step (S-3); A screen division step (S-3-2) of dividing the acquired image information into a plurality of image types to fit a frame, and a color difference signal division step (S-3-2) of obtaining color difference information of the divided image, and A screen division step (S-3-3) for each color difference signal divided according to the divided color difference signal, a pixel elementation step (3-3-4) for obtaining pixel values of each of the color difference signal screens, and the pixel value Applying the Markov chain for, loading the color difference signal information from the database (S-3-6), and comparing the color difference signal information of the database with the color difference signal information to which the Markov chain is applied (S-3) -7) ;.

The screen division step (S-3-2); divides the image information into an image form for each frame, and in this process, the image resolution is provided in QCIF or CIF. Then, the screen is divided according to each color difference signal of the RGB or CMYK method through the color difference signal division step (S-3-2); and the screen division step for each color difference signal (S-3-3).

In the pixel elementalization step (3-3-4); it is a step of displaying a screen divided according to a color difference signal as element values corresponding to each color. Then, in the step of applying the Markov Chain to the pixels (S-3-5); the Markov Chain described above is applied to each pixel.

Then, in the step (S-3-6) of loading the chrominance signal information from the database; the chrominance signal information of the image corresponding to the original artwork previously stored in the database is retrieved, and the chrominance signal information is compared (S- In 3-7) ;, the color difference signal information of the database and the color difference signal information to which the Markov chain is applied are matched and compared with each other.

In addition, the present invention is a step (S-3-7) of matching the color difference signal information; then, differentiating each unit process using a specific function (S-3-8) and the differentiating step. It is preferable to further apply the step of applying multiple layers (S-3-9) and the step of correcting the color difference signal information of the object to be differentiated through the error back propagation algorithm (S-3-10).

Here, the step of differentiating each unit process (S-3-8); may be, for example, a method of applying a Sigmoid function or an Arctan function, in the step of applying the multiple layers (S-3-9). The error back propagation algorithm is used in the step of correcting the color difference signal information (S-3-10) by sending the feedback to the step (S-3-7) of comparing the color difference signal information with various information again. The error backpropagation algorithm evaluates the differential value of the error function and sends it in the reverse direction of the progression to correct the color difference signal, so that the color difference information of the object to be differentiated is analyzed by correction learning by deep learning. It has the effect of improving the reliability.

In addition, the entropy type analysis step (S-4); a screen division step (S-4-1) of dividing the acquired image information into a plurality of image types according to a frame, as shown in FIG. 4, and the divided image Chrominance signal conversion step (S-4-2) for converting from RGB form to YUV form, and step (S-4-3) for passing the converted image through a fast discrete cosine transform (FDCT) process (S-4-3). Quantizing in bit form (S-4-4), statistically encoding the quantized data (S-4-5), and loading statistical code information for entropy from the database It consists of (S-4-6) and the step (S-4-7) of comparing the statistical code information of the database and the statistical coded data of the object to be discriminated using a Markov chain (S-4-7).

As described above, the screen division step (S-4-1) is a process of dividing the image information into an image form for each frame at the resolution of QCIF or CIF, and the color difference signal conversion step (S-4-2) is The color difference signal in RGB format is converted into YUV format. The YUV type signal is a method of expressing colors with a luminance signal (Y) and a color difference signal (U, V), and includes a luminance signal as well as a color difference signal.

The step (S-4-3) of performing the fast discrete cosine transform (FDCT) process is a method widely used for signal processing and image processing, and is a process for quantization and encoding of entropy. Then, in the step of quantizing (S-4-4); a quantization process is performed in the form of a bit as a result value through the conversion process. When this process is performed, binary data such as 0 or 1 is obtained.

The step of statistically encoding (S-4-5); is to undergo a statistical coding process using quantized data. Then, the step of retrieving the statistical code information (S-4-6) and the step of matching the pattern using the Markov chain with the statistical coded data (S-4-7); for the entropy of the divided spatial coordinate region. The pattern between the statistical code value and the statistical code value of the original artwork existing in the database is compared.

Then, determining the number of dimensions N and then comparing statistically encoded data in a multi-dimensional (N-dimensional) array (S-4-8) and collecting the multi-dimensional comparison results (S-4-9); By further processing, statistically encoded data on entropy as a whole is compared.

In the above, the present invention has been described with reference to the above embodiments, but it is of course possible to perform various modifications within the technical scope of the present invention.

10: database unit
20: image extraction unit
30: color type analysis unit
40: entropy type analysis unit
50: differential calculation unit
S-1 ~ S-5: Differentiation method for forgery of art using Markov chain

Claims (8)

delete delete delete A step (S-1) in which the image file of the original artwork is stored and statistical code information about the color difference signal information and the entropy of the entire original image is built in a database;
Step (S-2) of acquiring the image information of the artwork corresponding to the object of discrimination through the imaging equipment,
A color type analysis step (S-3) of dividing each image of the acquired image information according to a color difference signal and then applying a Markov chain to each pixel to match the color difference information stored in the database;
An entropy type analysis step (S-4) in which each image of the acquired image information is quantized and statistically coded for entropy and then compared with the statistical code information of the database in a Markov chain method;
Comprising of the color type analysis step (S-3) and the entropy type analysis step (S-4) by analyzing the correlation between the results of the calculation of the similarity with the original artwork (S-5);
The color type analysis step (S-3);
A screen division step (S-3-1) of dividing the acquired image information into a plurality of image types to fit the frame;
A chrominance signal segmentation step (S-3-2) of obtaining chrominance information of the segmented image;
A screen segmentation step for each color difference signal (S-3-3), which is divided according to the divided color difference signal;
A pixel elementalization step (3-3-4) of obtaining pixel values of each of the color difference signal screens;
Applying a Markov chain to the pixel (S-3-5),
Loading color difference signal information from the database (S-3-6),
Consisting of the color difference signal information of the database and the color difference signal information to which the Markov chain is applied (S-3-7); consists of,
The entropy type analysis step (S-4);
A screen division step (S-4-1) of dividing the acquired image information into a plurality of image types to fit the frame;
A color difference signal conversion step (S-4-2) of converting the segmented image from RGB to YUV;
Step (S-4-3) of passing the converted image through a fast discrete cosine transform (FDCT) process,
Quantizing in bit form as the result (S-4-4),
Statistically encoding the quantized data (S-4-5),
Step (S-4-6) of loading statistical code information for entropy from the database,
The step of comparing the statistical code information of the database and the statistically encoded data of the object to be discriminated using a Markov chain (S-4-7); Differentiating method for forgery and alteration of art using the Markov chain.
delete The method of claim 4,
(S-3-7) comparing the color difference signal information; then, differentiating each unit process using a specific function (S-3-8);
Applying multiple layers to the differentiating step (S-3-9);
A step of correcting color difference signal information of the object to be discriminated through an error back propagation algorithm (S-3-10); further comprising the method of discriminating forgery and alteration of art using a Markov chain. delete The method of claim 4,
After comparing the statistical code information of the database with the statistical coded data of the object to be discriminated using a Markov chain (S-4-7); after determining the number of dimensions N, and then in a multidimensional (N-dimensional) array In step (S-4-8) of comparing the statistically encoded data,
The step of collecting the multi-dimensional contrast results (S-4-9); discrimination method for forgery of art using a Markov chain further comprising.

Images