KR101549983B1 - The self-adaptive method for inserting/detecting a forensic mark using particle swarm optimization in the multiple transform domain and the apparatus thereof - Google Patents

Abstract

The present invention relates to a self-adaptive forensic mark inserting method using a particle swarm optimization method in a multiple transform domain, a detecting method, an inserting apparatus, and a detecting apparatus. The present invention provides the self-adaptive forensic mark inserting method using the particle swarm optimization method in the multiple transform domain which includes the steps of: (a) generating the forensic mark; (b) generating HH, LH, HL, and LL frequency domains through a 2D-discrete wavelet transform (2D-DWT); (c) drawing a final coefficient value DC; (d) drawing the optimization insertion position of the forensic mark by applying a particle swarm optimization (PSO) search method; (e) drawing an optimization insertion strength by the PSO search method; (f) quantizing the forensic mark according to the optimization insertion strength and inserting the quantized mark.

Description

Field of the Invention The present invention relates to a method for inserting a magnetic forensic mark into a multiple conversion domain using a particle cluster optimization technique, a detection method, an inserting device, and a detection device apparatus thereof}

The present invention relates to a forensic marking technique used for copyright protection and tracking of digital contents, and more particularly, to a method for robustly and securely inserting and detecting forensic marks in digital video contents using an optimized technique.

Digital contents can be copied indefinitely without loss of contents. Therefore, illegal copying and distribution are very likely on the Internet. In order to solve illegal copy / distribution of digital contents, illegal distribution of these contents on the Internet is fundamentally The copyright protection technology that blocks is becoming a topic of the times.

Previously, the watermarking technique was used to specify the copyright holder of the content. However, in this case, there is a problem that the same identification information is included in all the contents to be sold, making it difficult to trace the source of piracy. Forensic mark technology, which facilitates the tracking of illegal distributors of content that will occur later by inserting user information, is recognized as the most representative digital rights management technology.

Forensic Marking technology allows users to track and identify illegally duplicated users as it contains not only ownership information but also the information of the users who are being distributed. And it is a difficult technology to secure robustness and security. Therefore, forensic marking technology requires ① robustness ② invisibility ③ uniqueness ④ public acceptability ⑤ asymmetry ⑥ anonymity ⑦ conditional traceability.

Forensic Marks can be seen by applying the expected attack types to see if they have these technical elements: robust attack, open attack, move attack, upsampling, downsampling, filtering, clipping, D / A and A / D conversion, Attack types such as compression, picture rate variation, horizontal flip, shearing, linear geometric deformation, and complex deformation are typical.

At present, the forensic mark technique uses DWT (Discrete Wavelet Transform) or DCT (Discrete Cosine Transform), and Discrete Wavelet Transform (DWT) uses two or more filters in discrete signal processing, And then decomposing the frequency bands into different frequency bands. Unlike Fourier transform or discrete cosine transform, this transform method has frequency and time information at the same time and is useful in various fields. For two-dimensional images, the one-dimensional frequency decomposition process is generally performed in the horizontal and vertical directions. Since the two-dimensional image is down-sampled at a ratio of 2: 1 after passing through the low frequency and high frequency filters, The same before and after.

Discrete Cosine Transform (DCT) is a coefficient for transforming a video signal on a time axis into a frequency axis, and corresponds to an orthogonal transform coding method using a discrete cosine function. Especially, it is adopted in MPEG, which is an international standard for video compression, and is now becoming the mainstream of high-performance encoding and compression technology of video. In the discrete cosine transform, the video signal of the time axis is decomposed into a frequency domain and a small domain having several signal powers, and the power of the video signal is concentrated in the low frequency domain. Therefore, So that the data can be compressed. In the case of 3D discrete cosine transform, two-dimensional and one-dimensional DCT are performed on two-dimensional plane and time axis, respectively, and it is known to be effective for video compression.

However, the above-mentioned technology applied to the forensic marking field is focused on solving a single problem such as robustness or large capacity insertion because it is not easy to solve the problem of security as well as solving the problem of robustness, Considering performance is a very difficult problem. Therefore, there is a growing awareness of the necessity of research on technology development that enhances the robustness of forensic marks and improves security.

According to the present invention, it is an object of the present invention to provide a forensic marking method for robustly inserting / detecting a forensic mark safely generated in video contents by using a multiple conversion technique in order to protect digital contents, .

For this purpose, the present invention intends to apply a technology different from the watermarking technique using a single DWT or DCT, which has been mainly used in the prior art. In particular, a DFRNT transform is applied to a domain in which a forensic mark is subjected to 2D-DWT and 3D- (PSO search method), dither modulation (quantization), and so on, so that the robustness and security of the forensic mark can be improved while the detection performance can be improved while the invisibility is secured The present invention relates to a forensic marking method.

The present invention also provides a forensic marking method for improving security by providing a dual device for security of a forensic mark by constructing a security key using three parameters of DFRNT transformation and rotation order of a scrambling transformation The purpose.

In the present invention, 2D-DWT and 3D-DCT frequency conversion domain and DFRNT transform technique are used to secure robustness and security. PSO search technique (particle community optimization technique) and dither modulation dither modulation) quantization technique to construct a self - adaptive forensic marking.

The method includes the steps of: (a) generating a forensic mark based on insertion information to be inserted into digital image content data; (b) grouping the video frames of the digital image content data into a plurality of subframes, LH, HL and LL frequency bands through 2D-Discrete Wavelet Transform (2D-Discrete Wavelet Transform) transformation of the LH or HL frequency bands; (c) (DC) direct current by performing a DFRNT transformation (Discrete Fractional Random Transform) after transforming the final coefficient value into a DC-transformed 3D-Discrete Cosine Transform, Extracting the optimal insertion position of the forensic mark by applying a PSO search technique (Particle Swarm Optimization, Particle Swarm Optimization) to the target; (e) deriving an optimal insertion strength when the forensic mark is inserted at the optimum insertion position by a PSO search technique; and (f) quantizing and inserting the forensic mark according to the optimal insertion strength derived from the step (e) at the optimal insertion position derived from the step (d). And provides a self-adaptive forensic mark insertion method applied thereto.

A forensic mark insertion method and a forensic mark insertion / detection device are provided together with the insertion method.

When the digital image contents in which the forensic mark provided by the present invention is transferred through sale or the like, the robustness, non-visibility and security of the forensic mark are inserted at a level higher than that of the conventional forensic mark, And the detection performance of the inserted forensic mark is improved, so that the undamaged forensic mark can be detected.

These forensic marking technologies are effective in protecting content authors by allowing the broadcasters, movie studios, and online service providers to track their distributors when they distribute their video content and distribute them illegally on the Internet. to be.

In the present invention, since the original digital image contents data is converted into a frequency space to insert or extract a forensic mark, algorithms such as DWT transformation, DCT transformation, and DFRNT transformation are used to perform complex operations such as filtering and compression. Robustness and security can be improved.

In particular, the conventional homogeneous quantization is uniformly modified with both large and small coefficient values. On the other hand, according to the present invention, the small signal is corrected by the dither modulation quantization and the large signal is corrected by the dither modulation quantization, The performance and robustness can be improved.

FIG. 1 is a schematic flowchart of a method of inserting a magnetic adaptive forensic mark in which a particle cluster optimization technique is applied to multiple conversion domains according to the present invention.
FIG. 2 is a flowchart of a method for detecting a magnetically adaptive forensic mark in which a particle cluster optimization technique is applied to multiple conversion domains according to the present invention.
3 is a block diagram for deriving an optimal insertion position and insertion strength for insertion of a magnetically adaptive forensic mark in a magnetic adaptive forensic mark insertion method in which a particle cluster optimization technique is applied to multiple conversion domains according to the present invention .
4 is a flowchart illustrating a process of a PSO search method in a self-adaptive forensic mark insertion method in which a particle cluster optimization technique is applied to multiple conversion domains according to the present invention.
FIG. 5 is a block diagram of a method of inserting and detecting a magnetic adaptive forensic mark applying a particle cluster optimization technique to a multiple conversion domain according to the present invention.
FIG. 6 is a schematic diagram illustrating the principles of dither modulation quantization and inverse quantization in a magnetic adaptive forensic mark insertion method and a detection method in which a particle cluster optimization technique is applied to multiple conversion domains according to the present invention.
FIG. 7 is a block diagram of a magnetically adaptive forensic mark insertion detecting apparatus to which a particle cluster optimization technique is applied to multiple conversion domains according to the present invention.

Ⅰ. In the case of applying the particle cluster optimization technique to the multiple conversion domain according to the present invention Forensic Mark  Insert method

The method includes the steps of: (a) generating a forensic mark based on insertion information to be inserted into digital image content data; (b) grouping the video frames of the digital image content data into a plurality of subframes, LH, HL and LL frequency bands through 2D-Discrete Wavelet Transform (2D-Discrete Wavelet Transform) transformation of the LH or HL frequency bands; (c) (DC) direct current by performing a DFRNT transformation (Discrete Fractional Random Transform) after transforming the final coefficient value into a DC-transformed 3D-Discrete Cosine Transform, Extracting the optimal insertion position of the forensic mark by applying a PSO search technique (Particle Swarm Optimization, Particle Swarm Optimization) to the target; (e) deriving an optimal insertion strength when the forensic mark is inserted at the optimum insertion position by a PSO search technique; and (f) quantizing and inserting the forensic mark according to the optimal insertion strength derived from the step (e) at the optimal insertion position derived from the step (d). And provides a self-adaptive forensic mark insertion method applied thereto.

Further, the present invention further comprises that the rotation order of the three parameters used in the DFRNT transformation and the scrambling transformation in the step (c) is an insertion key or a detection key. It also provides a self-adaptive forensic mark insertion method applied.

The present invention also provides a PSO search method in (d) or (e), wherein (1) the specific coefficient values of at least one of the HH, LH, HL and LL frequency bands are particles Forming a group and initializing all particles in the group; (2) generating or updating a personal optimum value (pBest) or a cluster optimal value (gBest) by performing particle adaptability evaluation on each of the particles; And (3) if the number of particle adaptability evaluations reaches a set number of iterations, or if the difference between the individual optimal values and the cluster optimal values of the particles is less than a predetermined value, the search is terminated, Or the optimal insertion strength, and if not, repeating the step (2). The method for inserting a magnetic adaptive forensic mark according to an embodiment of the present invention, do.

Further, the present invention is characterized in that the particle adaptability evaluation is performed by calculating a robustness value (NC, Normalized Correlation) derived from an invisible value (PSNR) derived by searching for the insertion position of each particle and an attack type applied to the forensic mark The method comprising the steps of: (a) inputting a forensic mark to a plurality of domains;

(F) is a case where the value of the quantization function Q (f (i)) according to the following equation (2) is equal to the value of the corresponding forensic mark sequence equation w (i) (i)) = w (i)) is modified by the expression (3) and the value of the quantization function Q (f (i) (I)) is different from the value of the insertion target coefficient value f (i) by the dither modulation quantization in which the insertion subject coefficient value f (i) is modified by [Expression 4] The forensic mark is inserted into the multiple conversion domain. The forensic mark is inserted into the multi-conversion domain.

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

The present invention relates to a method of inserting a forensic mark having characteristics in an algorithm that can enhance the robustness of the forensic mark and improve both invisibility and security.

FIG. 1 is a schematic flowchart of a method of inserting a magnetic adaptive forensic mark in which a particle cluster optimization technique is applied to multiple conversion domains according to the present invention. Hereinafter, the process of each step will be described in detail.

(a)

(a) is a step of generating a forensic mark based on insertion information to be inserted into the digital image content data. The main characteristic of the forensic mark is that it describes the attribution information of the work in the work. Therefore, the content of the insertion information is different for each copyrighted work, unlike the existing watermark, so that a separate forensic mark is generated for each image content data.

(b)

(b) includes grouping the video frames of the digital image contents data into a plurality of subframes and then performing a 2D-DWT (2D-Discrete Wavelet Transform) conversion (S1) to convert HH, LH, LL frequency band.

The digital video content data corresponding to the original video signal is divided into a plurality of groups through video frame grouping based on the intermediate frame. When 2D-DWT conversion is performed on the intermediate frame of each group, the frequency bands of HH, LH, HL and LL, That is, decomposed into four subbands. In the present invention, the LH and HL frequency bands are converted. However, in the PSO search, four frequency bands are combined so as to enhance the aggressive response. This image processing process is performed separately from the process of generating the forensic mark, and corresponds to a transposition process of constructing an environment in which a forensic mark can be inserted by multiplexing the original digital image contents data.

In this conversion process, multiple conversion domains are formed. In FIG. 3, an 8 × 8 block DCT transformation is included in the urbanization.

(c)

(c) comprises performing a 3D-DCT (3D-Discrete Cosine Transform) transformation on the specific coefficient values in the LH or HL frequency band (S2) and then performing a DFRNT transformation (Discrete Fractional Random Transform) to extract the final coefficient value.

 The reason for selecting the HL or LH frequency bands out of the four frequency bands resulting from the 2D-DWT conversion is that the bands are highly robust against attacks against the forensic mark. Generally, the watermark is inserted into the LL frequency band which is most similar to the original image data, but the frequency band of HL or LH is selected to increase robustness.

HL and LH can be selected. If the insertion information amount is large, both of the two bands can be converted.

The 3D-DCT transform is performed on the selected LH or HL frequency band. The 2D-DCT transform is performed on the selected LH or HL frequency bands, and then the one-dimensional discrete cosine transform is performed again on a time axis frame basis. The coefficient values generated by the dual frequency conversion are input to the following DFRNT transform (S3) (Discrete Fractional Random Transform), randomly scrambled through parameter adjustment, and converted into unpredictable values.

DFRNT (Discrete Fractional Random Transform) is performed in the following manner. First, Matrix H, which is expressed by Equation 1 below, is generated using P generated as a random seed value, which is one of parameters of DFRNT .

In order to generate an eigenvector (eigenvector) from the matrix H, an SVD (Singular Value Decomposition) matrix decomposition is performed on H as in Equation (2).

은 [수학식 3]과 같이 N개의 orthogonal eigenvector로 구성된 matrix 이다.
At this time,

Is a matrix composed of N orthogonal eigenvectors as in Equation (3).

와 m을 이용하여 [수학식 4]와 같이 NⅹN의 diagonal matrix

를 생성한다.
Another parameter of the next DFRNT

And m are used to calculate the diagonal matrix of N x N as shown in Equation (4)

.

과

를 이용하여 [수학식 5]에 따라

를 계산하며 다시 계산된

와 DFRNT의 입력 신호인 X를 [수학식 6]에 대입하여 DFRNT 변환의 최종 출력인

를 얻는다.
And

and

According to Equation (5)

Lt; RTI ID = 0.0 >

And the input signal X of the DFRNT into the equation (6) to obtain the final output of the DFRNT transform

.

And further configuring the insertion key and the detection key as rotation orders of the three parameters and the scrambling transformation in the DFRNT transformation. The DFRNT transform can change the input signal into an arbitrary unpredictable signal with three parameters, which increases the computational complexity so that statistical characteristics can not be grasped, thereby improving the security of the forensic mark. Especially, the security can be improved by scrambling by scramble conversion. The scramble conversion is a conventional technique, and a description thereof will be omitted in the present invention.

The insertion key constructed as described above can be used as an insertion key when inserting a forensic mark and as a detection key for detecting a forensic mark.

On the other hand, since the transformed random signal can be reconstructed through inverse transform, it has a restoring force and improves the detection performance of the forensic mark.

In the present invention, since the original digital image contents data is converted into a frequency space to insert or extract a forensic mark, algorithms such as DWT transformation, DCT transformation, and DFRNT transformation are used to perform complex operations such as filtering and compression. Robustness and security can be improved.

(d) and (e)

(d) is a step of deriving an optimal insertion position of the forensic mark by applying a PSO search technique (Particle Swarm Optimization) to the domain of the final coefficient value, and step (e) Technique to derive the optimal insertion strength of the forensic mark to be inserted at the optimum insertion position.

3 is a block diagram for deriving an optimal insertion position and insertion strength for insertion of a magnetically adaptive forensic mark in a magnetic adaptive forensic mark insertion method in which a particle cluster optimization technique is applied to multiple conversion domains according to the present invention And FIG. 4 is a flowchart illustrating a process of a PSO search method in a method of inserting a magnetic adaptive forensic mark applying a particle cluster optimization technique to multiple conversion domains according to the present invention.

The steps (d) and (e) correspond to searching for the optimal insertion position and insertion strength of the forensic mark by the PSO search technique. The insertion position and insertion strength of the forensic mark are searched using the PSO search technique The robustness and detection performance are enhanced. In order to determine the optimal insertion position, it is desirable to use PSO search method once again to determine the optimal insertion strength after applying the PSO search method in order to secure invisibility and robustness.

Referring to FIGS. 3 and 4, Particle Swarm Optimization (PSO), or PSO search, is one of the evolutionary techniques. As with other evolutionary techniques, Start by setting up groups. However, in PSO search technique, each of these arbitrary solutions is referred to as a particle, and in order to find an optimal solution by changing the particles to an optimized value, unlike the existing genetic techniques, By referring to the information about the optimal solution of the surroundings, it changes the velocity of the user and selects the strategy of changing the position, that is, the particle value. So as the optimization progresses, each particle seems to fly in an area that is closer to the optimal location than where it is now.

In the present invention, such a PSO search technique is utilized to find the optimal position to be inserted for the forensic mark, and at the same time, to determine the best insertion strength at the optimum insertion position.

The PSO search method for extracting the insertion position and the insertion strength of the forensic mark can perform the following process step by step.

(Stage 1

In the step (1), a group consisting of specific coefficient values of one or more frequency bands of the HH, LH, HL and LL frequency bands is formed and all particles in the group are initialized. Corresponds to the process from S4-11 to S4-13 in FIG. 3.

In the present invention, an initial value is set for one or more frequency bands of the LL, HL, and LH frequency bands in which the digital image content data corresponding to the original is converted, and a process of deriving and analyzing an optimal value at each position is performed . In this case, the initial value may be a signal that has been subjected to 2D-DWT conversion, 3D-DCT conversion, or DFRNT conversion in the original digital image contents data.

To do this, a specific coefficient value of a frequency band converted into multiple frequencies is first selected as a particle to form a group, and all particles in the group are initialized so that their positions and velocities are zero.

(2) Step

(2) is a step of generating or updating a personal optimum value (pBest) or a cluster optimal value (gBest) by performing particle adaptability evaluation for each particle performance. Corresponds to the process of S4-14 and S4-15 in FIG. 3.

The PSO search method searches for the solution of the optimal value by comparing with the surrounding value instead of evolving through the specific operation. Therefore, the individual optimum value (pBest) or the cluster optimal value (gBest) And the performance of the particle performance is evaluated to update the individual optimum value and the cluster optimal value to the current optimal value. In this case, the individual optimum value or the cluster optimum value may be derived from the final count value.

The individual optimum value or the cluster optimal value may be an insertion position or an insertion strength according to a step. For example, if the PSO search method is used in step (d), the value of the insertion position is input, and in step (e), the insertion strength value is input or calculated.

The performance evaluation can be performed for the evaluation of particle adaptability. At this time, according to the nonvisibility value (PSNR) of the forensic mark and the attack type applied to the forensic mark, the robustness value (NC, Normalized Correlation) Is calculated by the performance evaluation function ([Expression 1]).

[Equation 1]

PSNR: Invisibility value

NC: Robustness value

m: Type of attack used in PSO technique

This performance evaluation function is based on the non-visibility (PSNR) value and the robustness (NC) value, and is used to search the insertion strength after determining the insertion position by the PSO search technique. Normalized Correlation (NC) is a normalized correlation coefficient. If no error is generated at all, the value is 1, and if all errors occur, the value is zero. That is, if the detected forensic mark and the original forensic mark are completely matched, the NC value becomes 1. If the forensic mark is detected after the content in which the forensic mark is inserted is subjected to various attacks, the NC value is smaller than 1 and larger than 0 .

The PSNR (Peak Signal to Noise Ratio) is used as an index for evaluating the quality of a video signal with a similar SNR to the maximum signal-to-noise ratio (SNR). For the insertion or detection of a forensic mark, it is desirable to secure this value at least 40 dB.

(M) of the low-frequency filter attack, noise attack, etc. used in the PSO search method of the insertion position. Attack types include robustness attack, open attack, mobile attack, upsampling, downsampling, filtering, clipping, D / A and A / D conversion, compression, aspect ratio transformation, horizontal flip, shearing, linear geometry, And other types of attacks.

For example, if an initial value is set for a specific coefficient value in the LL frequency band, a low pass filtering attack is searched as a model to derive an NC value. For the HL or LH frequency band, a noise attack or a compression attack is modeled The NC value is derived.

If the PSNR value is set to 40 dB, the optimum insertion position or optimal insertion strength is obtained when the fitness value calculated using the NC value derived from the attack type selected in the frequency band is the maximum.

(3) Step

In step (3), if the number of particle adaptability evaluations reaches a set number of repetitions, or if the difference between the individual optimal value of the particle and the optimal cluster value is less than a predetermined value, the search is terminated, Position or the optimal insertion strength, and if not, performing the step (2) again. Corresponds to the process of S4-16 and S4-17 in FIG. 3.

In general, when the number of repetitions of the number of particle adaptability evaluations is set in advance, the particle value is estimated as the final value when a certain number of times are repeated. Further, the difference between the individual optimum value or the cluster optimal value and the corresponding particle value is set in advance If the value is less than the preset value, the particle value is regarded as the final value and the search is stopped.

In this way, unlike the existing technique, the PSO does not exchange information of its own information and arbitrary neighboring particles but refers to the information of the particles which are considered to be close to the most optimal solution in the cluster, I can find the solution soon.

In the present invention, the PSO search technique is applied to derive the optimum insertion position and optimal insertion strength. First, the optimal insertion position of the forensic mark is derived by the PSO search technique, and then the optimal insertion strength is derived by applying the PSO search technique .

When the optimal insertion position or the optimal insertion strength is derived by the PSO search technique, it is possible to examine whether the optimal insertion position or the optimal insertion strength is correct as a performance evaluation method used in the particle adaptability evaluation to evaluate its performance. In particular, when the forensic mark is inserted, PSNR and NC are calculated to check the invisibility and robustness by the forensic mark algorithm. In this case, not the fitness value in [Equation 1] is calculated but the non-visibility (PSNR) NC) to check.

(f)

(f) is a step of quantizing and inserting the forensic mark according to the optimal insertion strength derived from the step (e) at the optimum insertion position derived from the step (d). According to the PSO search technique, the forensic mark can be quantized and inserted into the optimized insertion position at an optimal insertion strength.

FIG. 6 is a schematic diagram illustrating the principles of dither modulation quantization and inverse quantization in a magnetic adaptive forensic mark insertion method and a detection method in which a particle cluster optimization technique is applied to multiple conversion domains according to the present invention. Referring to FIG. 6, the dither modulation quantization can be confirmed.

In particular, the present invention is applicable to the case where the value of the quantization function Q (f (i)) according to the following equation (2) is equal to the value of the sequence formula w (i) (i)) of the forensic mark and the value of the quantization function Q (f (i)) is corrected by the value of the sequence formula w (i) of the forensic mark, In another case (Q (f (i)) w (i)), the forensic mark is obtained through dither modulation quantization in which the insertion subject coefficient value f (i) The method of inserting a magnetic adaptive forensic mark according to an embodiment of the present invention provides a method of inserting a magnetic adaptive forensic mark using a particle cluster optimization technique in a multiple conversion domain.

4 is a schematic diagram illustrating the principles of dither modulation quantization and inverse quantization in a magnetic adaptive forensic mark insertion method and a detection method in which a particle cluster optimization technique is applied to multiple conversion domains according to the present invention. Referring to this, w (i) represents that the forensic mark itself is a sequence, and means that individual elements of this sequence are quantized and inserted into the frequency coefficient value. i corresponds to the index of the sequence element.

[Equation 2]

Q: Quantization function

f: Insertion target coefficient value

: 양자화 강도를 나타내는 양수값

: A positive value indicating the quantization strength

[Equation 3]

f (i): insertion target coefficient value, i = 1, 2, 3 .... M × M

: 양자화 강도를 나타내는 양수값

: A positive value indicating the quantization strength

[Equation 4]

f (i): insertion target coefficient value, i = 1, 2, 3 .... M × M

: 양자화 강도를 나타내는 양수값

: A positive value indicating the quantization strength

The reason for performing the quantization process according to the dither modulation in the present invention is that the general quantization process is uniformly modified in all values whether the value of the signal, that is, the target coefficient value is large or small, while in the case of the dither modulation quantization, This is because the signal is largely corrected so that the invisibility and robustness before and after insertion of the forensic mark are greatly enhanced.

After the quantization process, the final coefficient value (DC) in which the forensic mark is inserted is generated as the digital image content data, which is the combination of the original frames through the inverse DFRNT transformation, the inverse 3D-DCT transformation and the 2D-DWT transformation, And is circulated in an invisible state.

Ⅱ. In the case of applying the particle cluster optimization technique to the multiple conversion domain according to the present invention Forensic Mark detection method

(A) a method for grouping video frames of digital image contents data into which a forensic mark is inserted according to the forensic mark insertion method of the above (1), into a plurality of subframes and then performing a 2D-DWT (HL, HL, and LL) frequency bands of the LH or HL frequency bands through a 3D-DCT transform (Wavelet Transform) (3-D discrete cosine transform, 3D-Discrete Cosine Transform), and then performing a DFRNT transform (Discrete Fractional Random Transform) to extract a final DC value; And (C) detecting a forensic mark from the final coefficient value through inverse quantization. The present invention also provides a method for detecting a magnetic adaptive forensic mark in which a particle cluster optimization technique is applied to multiple conversion domains.

Also, the present invention provides a method for detecting a magnetic adaptive forensic mark using a particle cluster optimization technique in a multiple conversion domain, wherein the generating key generated in the step (C) is used as a detection key.

FIG. 2 is a flowchart of a method for detecting a magnetically adaptive forensic mark in which a particle cluster optimization technique is applied to multiple conversion domains according to the present invention. The forensic mark detection method can be regarded as a reverse process of the forensic mark insertion method of the present invention. However, as can be seen from FIG. 2, the time series can be derived through the same process as the forensic mark insertion method.

The steps (A) and (B) may be considered to be the same as steps (a) and (b) of the forensic mark insertion method. (C) relates to a process of detecting a forensic mark embedded in digital image contents data through inverse quantization (S6). In particular, inverse quantization of dither modulation can be used as an inverse quantization method, which can be expressed by the following equation.

In Equation (7), mod is a meaning of modulation, which corresponds to an equation for dividing a certain number by 2 and obtaining the remainder.

In particular, in the present invention, The generation key can be used as a detection key. In order to improve the detection performance of the forensic mark and to improve the security, the detection key must be the same as the generation key.

The process of inserting and detecting a forensic mark according to the present invention is a block diagram of a method for inserting and detecting a magnetic adaptive forensic mark applying a particle cluster optimization technique to a multiple conversion domain according to the present invention of FIG. Can be confirmed.

Ⅲ. Applying Particle Swarm Optimization Techniques to Multiple Conversion Domains Forensic Mark  Insertion / detection device

A forensic mark generation unit (100) for generating a forensic mark based on insertion information to be inserted into digital image content data; An image processing unit 200 for performing a 2D-DWT conversion, a 3D-DCT conversion, and a DFRNT conversion sequentially on the digital image contents data to generate specific coefficient values; An insertion detecting unit 300 for deriving an optimal insertion position and an optimal insertion strength of a forensic mark inserted into the specific coefficient value by a PSO technique and quantizing and inserting the forensic mark or dequantizing the forensic mark from the specific coefficient value The present invention provides a self-adaptive forensic mark inserting and detecting apparatus that applies a particle cluster optimization technique to multiple conversion domains including a plurality of conversion domains.

The forensic mark inserting / detecting apparatus of the present invention is hardware in which the forensic mark inserting method and the inspecting method of the present invention described above are performed, and the form and arrangement of the forensic mark inserting / detecting apparatus are manufactured in various ways .

FIG. 7 is a block diagram of a magnetically adaptive forensic mark insertion detecting apparatus to which a particle cluster optimization technique is applied to multiple conversion domains according to the present invention. Referring to this, the forensic mark generation unit 100 plays a role of generating a forensic mark based on insertion information to be inserted into the digital image content data. The forensic mark is generated and stored for each digital image content data because individual information is received according to the seller and the purchaser. The forensic mark generation unit 100 encrypts the insertion information to generate a forensic mark, stores the forensic mark and stores the forensic mark until the insertion position and the insertion strength are determined after the image processing for the digital image contents data is performed .

After the forensic mark is generated, a role of performing multiple conversion on the digital image contents data to insert the forensic mark is performed in the image processing unit 200. The image processing unit 200 performs steps (b) to (e) or (B) to (C) of the insertion method and detection method of the present invention described above.

In the image processing unit, 2D-DWT transformation, 3D-DCT transformation, and DFRNT transformation are successively performed on the digital image contents data to form a domain through multiple transforms to form multiple transform domains.

The insertion detecting unit 300 may derive the optimal insertion position and the optimal insertion strength of the forensic mark inserted into the specific coefficient value using the PSO technique to quantize and insert the forensic mark or dequantize the forensic mark from the specific coefficient value .

The forensic mark is inserted in the multiple conversion domain derived from the image processing unit We use the PSO search method to derive the location and strength. The PSO search technique may proceed to the steps (1) to (3) of the insertion method of the present invention described above.

The forensic mark generating unit, the image processing unit, and the insertion detecting unit may be physically separated into one organic structure.

While the present invention has been particularly shown and described with reference to exemplary embodiments thereof, it is to be understood that the invention is not limited to the disclosed embodiments, but, on the contrary, is intended to cover various modifications and equivalent arrangements included within the spirit and scope of the invention. Accordingly, the claims of the present invention include modifications and variations that fall within the true scope of the present invention.

100: Forensic Mark Generation Unit 200:
300: insertion detector

Claims (8)

(a) generating a forensic mark based on insertion information to be inserted into the digital image content data;
(b) generating HH, LH, HL and LL frequency bands through 2D-DWT (2D Discrete Wavelet Transform) conversion after grouping the video frames of the digital image contents data into a plurality of subframes ;
(c) 3D-DCT (3D-Discrete Cosine Transform) conversion of specific coefficient values in the LH or HL frequency band, and then DFRNT transformation (discrete fractional random transformation) Deriving a final count value (DC, direct current);
(d) deriving an optimal insertion position of the forensic mark by applying a PSO search technique (Particle Swarm Optimization) to the final coefficient value;
(e) deriving an optimal insertion strength when the forensic mark is inserted at the optimum insertion position by a PSO search technique;
(f) quantizing and inserting the forensic mark according to the optimal insertion strength derived from the step (e) at the optimum insertion position derived from the step (d);
A method of inserting a magnetic adaptive forensic mark applying particle group optimization method to multiple transform domain including.
The method of claim 1,
The step (c)
Further comprising that the rotation order of the scrambling transformation and the three parameters used in the DFRNT conversion serve as an insertion key or a detection key. ≪ RTI ID = 0.0 > [10] < / RTI >
The method of claim 1,
In the step (d) or the step (e)
(1) forming a group of particles having specific coefficient values of at least one frequency band among the HH, LH, HL and LL frequency bands and initializing all the particles in the group;
(2) generating or updating a personal optimum value (pBest) or a cluster optimal value (gBest) by performing particle adaptability evaluation on each of the particles; And
(3) if the number of particle adaptability evaluations reaches a set number of repetitions, or if the difference between the individual optimal value and the cluster optimal value of the particle is less than a preset value, the search is terminated and the individual optimal value is set to the optimum insertion position or Setting the optimal insertion strength to the optimum insertion strength; otherwise, performing the step (2) again;
The method of claim 1, wherein the method further comprises:
4. The method of claim 3,
The particle adaptability evaluation may be performed,
(PSNR) obtained by exploring the insertion position of each particle and a robustness value (NC, Normalized Correlation) derived from the type of attack applied to the forensic mark are used as the performance evaluation function of [Equation 1] The method comprising the steps of: (a) inputting a plurality of transformed domains;
[Equation 1]

PSNR: Invisibility value
NC: Robustness value
m: Type of attack used in PSO technique
The method of claim 1,
The step (f)
When the value of the quantization function Q (f (i)) according to the following equation (2) is equal to the value of the sequence formula w (i) of the corresponding forensic mark (Q (f (i)) = w (i) When the value of the quantization function Q (f (i)) is different from the value of the sequence formula w (i) of the forensic mark Q (i) the forensic mark is inserted through dither modulation quantization in which the insertion subject coefficient value f (i) is modified by the following equation (4): (f (i) A method of inserting a magnetic adaptive forensic mark applying particle group optimization method to multiple transform domain.
[Equation 2]

Q: Quantization function
f: Insertion target coefficient value

: A positive value indicating the quantization strength

[Equation 3]

f (i): insertion target coefficient value, i = 1, 2, 3 .... M × M

: A positive value indicating the quantization strength

[Equation 4]

f (i): insertion target coefficient value, i = 1, 2, 3 .... M × M

: A positive value indicating the quantization strength
(A) a method of grouping video frames of digital image contents data into which a forensic mark is inserted according to the forensic mark insertion method of claim 1, into a plurality of sub-frames and then performing 2D-DWT (2D-Discrete Wavelet Transform) Generating the HH, LH, HL, and LL frequency bands through the frequency domain;
(B) performing 3D-DCT (3D-Discrete Cosine Transform) on the specific coefficient values of the frequency band selected in the insertion method of the first aspect among the LH or HL frequency bands and performing DFRNT transformation Performing a discrete Fractional Random Transform to extract a final DC value; And
(C) detecting a forensic mark from the final coefficient value through inverse quantization using the generation key generated in the third step as a detection key;
A method for detecting a magnetic adaptive forensic mark applying a particle cluster optimization technique to a multiple transform domain including a plurality of transform domains.
6. An insertion device in which the insertion method of any one of claims 1 to 5 is performed,
A forensic mark generation unit (100) for generating a forensic mark based on insertion information to be inserted into the digital image content data;
An image processor 200 for successively performing 2D-DWT transformation, 3D-DCT transformation, and DFRNT transformation on the digital image contents data to generate an encrypted final coefficient value DC;
An insertion detecting unit (300) for deriving an optimal insertion position and an optimal insertion strength of a forensic mark inserted into the final count value by a PSO technique and quantizing and inserting the forensic mark;
A method for inserting a magnetic adaptive forensic mark into a plurality of transform domains including a particle group optimization technique.
7. A detection device in which the detection method of claim 6 is performed,
A forensic mark generation unit (100) for generating a forensic mark based on insertion information to be inserted into the digital image content data;
An image processor 200 for successively performing 2D-DWT transformation, 3D-DCT transformation, and DFRNT transformation on the digital image contents data to generate an encrypted final coefficient value DC;
An optimal insertion position and an optimal insertion strength of the forensic mark inserted into the final count value are derived by the PSO technique and the forensic mark is quantized to dequantize the forensic mark inserted in the digital image content data from the specific coefficient value An insertion detecting unit 300;
And a particle group optimization technique is applied to the multiple conversion domain including a plurality of conversion domains.

Images