KR100896618B1 - Apparatus and method for inserting and detecting digital image watermarking - Google Patents

Abstract

The present invention relates to a watermark embedding / detection apparatus and a method thereof.

According to the present invention, when an image is subjected to rotation, scaling, or translational geometrical deformation, the image is subjected to 2D FFT and then takes an amplitude, and the 2D FFT amplitude is rotated at the same angle, and is deformed in inverse proportion to the scaling. Using an unaffected movement, a watermark derived from a sequence of pseudonoises is inserted in a particular area above the straight lines passing through the origin in the 2D FFT amplitude area. And, even though the geometric deformation occurs in the image, the points of the straight lines passing through the origin in the 2D FFT amplitude region are rotated equally, and the angle and the scaling are determined using a distance from the origin in inverse proportion to the scaling. The watermark detection area is reset and detection is performed using a whitening filter to find the watermark.

Similar Noise Sequence, White Filter, Amplitude, 2D FFT, Watermark

Description

Apparatus and method for inserting and detecting digital image watermarking}

1 is a block diagram of a watermark embedding apparatus according to an embodiment of the present invention.

2 illustrates a watermark embedding process according to an embodiment of the present invention.

3 illustrates a 2D FFT transformed region of a digital image according to an exemplary embodiment of the present invention.

4 is a block diagram of a watermark detection apparatus according to an embodiment of the present invention.

5 is a diagram illustrating a watermark detection process according to an embodiment of the present invention.

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to watermarking, and more particularly, an insertion / detection apparatus and method for inserting / detecting a watermark, which is robust against geometric deformations such as rotation, scaling, movement, and the like, into a digital image signal, and the above-described method. The present invention relates to a computer-readable recording medium having recorded thereon a program.

With the development of information and communication technology, various data forms are changed from analog form to digital form, which is easy to process and does not cause any quality deterioration even after playing and copying many times. It is widely distributed and distributed.

Among them, digital images are widely used on the Internet due to the development of compression techniques such as joint photographic coding experts group (JPEG).

Such digital images are easily copied and distributed, and thus, when the digital images are distributed indiscriminately in an open space such as the Internet without the author's consent, the intellectual property of the original author and the owner is not protected.

In order to solve this problem, a method of embedding non-cognitive signals in digital images and using them for authentication and protection of ownership is a watermarking method that can detect and claim forgery, alteration, and illegal use of digital images. It is proposed.

There are various watermarking methods developed at present, but nowadays, a method using a pseudo random sequence is mainly used. In addition, a transformed domain using a Discrete Fourier Transform (DFT) or a Discrete Cosine Transform (DCT) or the like is mainly used for a region in which a watermark is inserted.

Most algorithms using similar noise sequences detect a watermark inserted into a similar noise sequence using a matched filter. Most watermark detection methods currently use a blind detection method that detects an original image without an original image.                         

However, in the blind detection method, when the digital image is geometrically deformed such as rotation, scaling, and parallel movement by the user, it is impossible to detect the watermark because there is a problem in hydrothermal synchronization during the pseudo-noise sequence test. In addition, when the watermark is detected, there is a problem in that the detection performance is degraded because the original video signal acts as a noise.

SUMMARY OF THE INVENTION The present invention has been made to solve the above-described problems, and an object of the present invention is to insert a watermark, which is robust against geometrical deformations such as rotation, scaling, and parallel movement of a digital image, into a digital image and to easily extract the watermark. .

Watermark embedding and detection apparatus according to the present invention for achieving the above technical problem,

A pseudo-noise sequence generator that receives a user's key from the outside and generates a pseudo-noise sequence corresponding to the user's key, and a 2-dimensional fast Fourier transform by zero padding a digital image (2D FFT: 2 domain Fast Fourier Transform) A Fourier transform unit for performing the following steps, a component separator for dividing the amplitude component and the phase component in the two-dimensional fast Fourier transform digital image, and a straight line including the amplitude components passing through the origin in the amplitude region obtained by the component separator. An insertion region determiner for selecting and determining a point corresponding to the user key as a watermark insertion point, a watermark generator for generating a watermark using the pseudonome sequence generated by the pseudonome sequence; Embed the watermark generated by the watermark generator so as to match the point determined by the insertion area determiner. A watermark embedding unit and a watermark embedding digital image through a two-dimensional inverse fast Fourier transform and zero remove of an amplitude region and a phase component into which a workmark is inserted by the watermark inserting unit It is characterized in that it comprises a watermark insertion image generation unit.

In addition, the present invention according to another feature for achieving the above technical problem,

A pseudo-noise sequence generator that receives a user key from an external source and generates a pseudo-noise sequence corresponding to the user key, a Fourier transform unit that performs two-dimensional fast Fourier transform by zero-padding a digital image, and the two-dimensional fast Fourier transform A component separation unit for dividing an amplitude component and a phase component in a digital image, and a straight line including amplitude components passing through an origin in an amplitude region obtained by the component separation unit, and selecting a straight line, An extraction region determiner for determining matching points as a watermark extraction point, a watermark generator for generating a watermark using the pseudonome sequence generated by the pseudonome sequence, and the watermark generator the sequence of the watermark is generated by (W _n) and the watermark estimation provided by said extraction region determiner Characterized by including the point of the sequence (J _n) for use by a correlator for calculating a correlation value, and, to determine the watermark to determine a correlation value and set whether the watermark detected using the threshold value of the correlator unit.

In addition, the present invention according to another feature for achieving the above technical problem,

A first step of generating a similar noise sequence using a user key input from the outside; a second step of zero padding the original signal to be inserted into the watermark input from the outside; and a second step of the two-dimensional fast Fourier transform key; A third step of separating the components into amplitude and phase components in a Fourier transformed digital image, and selecting a straight line including amplitude components passing through the origin in the amplitude region obtained by the component separating unit, A fourth step of determining a point corresponding to the user key as a watermark insertion point, a fifth step of generating a watermark corresponding to the similar noise sequence and inserting the watermark into the insertion point determined in the fourth step; And a sixth step of making the watermark-embedded amplitude region into a zero-eliminated digital image. And a gong.

In addition, the present invention according to another feature for achieving the above technical problem,

A first step of receiving a user key from an external source and generating a similar noise sequence corresponding to the user key, a second step of zero-padding a digital image into which a watermark is input from an external source and inserting a two-dimensional fast Fourier transform; A third step of dividing the two-dimensional fast Fourier transformed digital image into an amplitude component and a phase component, and selecting a straight line including amplitude components passing through the origin in the amplitude region obtained by the component separation unit, and a point located on the selected straight line A fourth step of determining points J _n corresponding to the user key as a watermark extraction point, a sequence of watermarks W _n generated by the watermark generator, and the extraction area determination unit and a fifth step of calculating a correlation value using a random number sequence (J _n) of the watermark extraction point provided by, take advantage of the correlation values and the set door teokgap Characterized in that a sixth step of determining whether the watermark is detected.

Hereinafter, a watermark embedding and detecting apparatus and method thereof according to an embodiment of the present invention will be described with reference to the accompanying drawings.

1 is a block diagram of a watermark embedding apparatus according to an embodiment of the present invention. As shown in FIG. 1, the watermark embedding apparatus according to an embodiment of the present invention includes a Fourier transform unit 11, a pseudo-noise sequence generator 12, a component separator 13, and an insertion region determiner 14. And a watermark generator 15, a watermark inserter 16, and a watermark embedded image generator 17.

The present invention is characterized in that a watermark is inserted in an amplitude region that is robust to geometric deformation in a 2D FFT region of a digital image.

Accordingly, the Fourier transform unit 11 is for converting a digital image into a 2D FFT, and the component separation unit 13 is for separating phase and amplitude components from the 2D FFT digital image. ) Is for determining the area where the watermark is to be inserted. In this case, the insertion region determiner 14 uses a user key and an amplitude component of the digital image, and also uses a straight line that includes the amplitude component and passes through the origin.

The similar noise sequence generation unit 12 receives a user key containing ownership information and generates a similar noise sequence corresponding to the user key. Since the pseudo-noise sequence has a statistically almost irregular characteristic, it is impossible to trace back when the user key is not known, but when the user key is known, it can be reproduced through a mechanical process. Therefore, the pseudo noise sequence generated by the pseudo noise sequence generator 12 becomes a pseudo noise sequence that only a user can decipher.

The watermark generating unit 15 generates a watermark corresponding to the similar noise sequence generated by the similar noise sequence generating unit 12. The watermark inserting unit 16 receives the outputs of the insertion region determining unit 13 and the watermark generating unit 14 and inserts the watermark into the amplitude component of the digital image. The watermark embedding image generating unit 17 adds the amplitude component into which the watermark is inserted and the phase component separated by the component separating unit 13 to form a digital signal in a form that can be read by a computer or the like and inverse Fourier. Convert and perform zero remove to bring the original digital signal to specification.

Hereinafter, an operation of the watermark embedding apparatus shown in FIG. 1 will be described with reference to FIGS. 2 and 3.

2 is a view showing a watermark embedding process according to an embodiment of the present invention, Figure 3 is a view showing a 2D FFT transformed frequency domain of a digital image according to an embodiment of the present invention.

As shown in FIG. 2, when the original image i (x, y), which is a T × T digital image signal, is input, the Fourier transform unit 11 converts the original image i (x, y) to K (K is 2). , 3, 4, ....) T × KT size to zero padding (S210).

Here, the zero padding is to increase the resolution of the 2D FFT to be performed later. For example, when the size of the original image is 512 (width) x 512 (length), zero padding is performed at a much larger size than the original image, such as 1024 × 1024 or 2048 × 2048. In this case, it is preferable that the value of K be large enough so that the original image of the T × T size can be brought into the zero-padded size even if it is geometrically deformed such as rotation, scaling, or parallel movement by the operator. However, if the K value becomes too large and the size of the zero patting increases, the resolution of the 2D FFT increases, but the amount of calculation increases accordingly, and thus the K value needs to be adjusted.

In the present invention, when the original image is zero-patterned to a size of 2T × 2T as shown in FIG. 3, an embodiment will be described.

The zero-padded Fourier transform unit 11 as shown in FIG. 3 obtains an I (X, Y) value by performing two-dimensional fast Fourier transform (2D FFT) on the zero-padded i (x, y), and the component separation unit Output to (13) (S211).

Here, each element of I (X, Y) is composed of a complex number, the real part of the complex number is an amplitude component, and the imaginary part is a phase component.

로 표현되고, 위상 성분은 그대로 PI(X,Y)로 표현된다. 본 발명은 진폭 성분에 워터마크를 삽입하는 것을 특징으로 하므로, 성분 분리부(13)에서 분리된 진폭 성분(

)는 삽입영역 결정부(14)로 출력된다.When the component separation unit 13 inputs a 2D FFT video signal from the Fourier transform unit 11, the component separation unit 13 separates an amplitude component and a phase component (S212). At this time, the amplitude component is represented by an absolute value.

The phase component is expressed as PI (X, Y) as it is. The present invention is characterized in that a watermark is inserted into an amplitude component, so that the amplitude component separated by the component separation unit 13 (

) Is output to the insertion region determination section 14.

)를 입력받아 진폭 성분에 워터마크가 삽입될 영역을 결정하게 되는데, 이때의 워터마크 삽입 영역은 사용자 키에 따라 결정하게 된다(S213).The insertion region determiner 14 receives the amplitude component (a) from the component separator 13.

), An area in which the watermark is to be inserted into the amplitude component is determined, and the watermark embedding area is determined according to the user's key (S213).

This is because the watermark generated by the user key only matches the user key. The insertion region determiner 14 should select an amplitude region that is less affected by rotation, reduction / enlargement, and movement of an image in determining the watermark insertion region. An amplitude region that is robust to such geometric deformation is shown in FIG. 3. the zero point crossing a straight line as shown in the _{(L 1, L 2, ...} , L m, ..., L m). At this time, the straight line is made by the amplitude component, and the straight line passing through the origin shows the shape of the straight line passing through the origin even after the geometric deformation, so that the detection is easy even after the geometric deformation.

Hereinafter, an operation of determining the watermark embedding area on the straight lines L ₁ , L ₂ , ..., L _m , ..., L _M of the amplitude area performed by the embedding area determiner 14 will be described.

For example, suppose that the original image i (x, y) is rotated by α in the spatial domain, the image is scaled horizontally and vertically by σ, and moved by (x0, y0).

When the image generated by the transformation of the RST (Rotation, Scaling, Translation) is j (x, y), j (x, y) can be expressed as Equation 1 below.

는 다음의 수학식 2와 같이 나타낼 수 있다. Then, after 2D FFT of j (x, y), the absolute value is taken.

Can be expressed as Equation 2 below.

배 만큼 커지지만 검출에 아무런 영향을 미치지 않는다. 또한, (x0, y0)만큼 평행이동한 것은

에 아무런 영향을 미치지 않는다.As shown in Equations 1 and 2, J (X, Y) is obtained by rotating I (X, Y) by α, and then scaling them horizontally and vertically by α ⁻¹ times. The values at all points

It is twice as big but does not affect detection. Also, the parallel movement by (x0, y0)

Does not affect anything.

In other words, the geometric deformation is proportional to the rotation, inversely proportional to the scaling, and the translation does not affect.

에서 원점을 지나는 직선 위의 점 중에서 워터마크가 삽입될 점의 집합인 {

}는 i(x, y)가 α만큼 회전되고, σ만큼 영상을 가로, 세로로 크기조정하게 되면, α만큼 회전하게 되고 1/σ배만큼 가로, 세로 좌표가 크기 조정된다.therefore,

The set of points on the straight line through the origin at which the watermark will be inserted {

} Rotates i (x, y) by α, scales the image horizontally and vertically by σ, rotates by α and scales the horizontal and vertical coordinates by 1 / σ times.

For this reason, the insertion region determination section 14 selects straight lines L ₁ , L ₂ , ..., L _m , ..., L _M passing through the origin. Then, a point to insert a watermark on each of the straight lines is selected. In this case, the low frequency region whose distance is relatively close to the origin, that is, the point whose radius is within R1 is sensitive to the image, is excluded from the insertion region. The high-frequency region, that is, the point whose radius is outside R2, is damaged during signal processing such as compression. Thus, as shown in FIG. 3, the distance from the origin is a point between R1 and R2 and above the straight line L ₁ , L ₂ , ..., L _m , ..., L _M passing through the origin. Select the points.

}이며, 따라서, 수열 {

}는 하기의 유사잡음수열(Un)과 워터마크(Wn)의 데이터 수와 동일한 {I₁, I₂, ..., I_N)가 된다.Since the points selected in this way are sampled by K at regular intervals for each straight line, the total number becomes M (number of straight lines) x K (sampling number). As shown in Fig. 3, five points k1, k2, k3, k4 and k5 are sampled on each straight line. Here, the sampled point, i.e., the area (i.e., the point) into which the watermark is to be inserted, is {

}, Thus, the sequence {

} Becomes {I ₁ , I ₂ ,..., I _N ) equal to the data numbers of the similar noise sequence Un and the watermark Wn described below.

On the other hand, the present invention is characterized by using a similar noise sequence when embedding the watermark in the digital image, it requires a similar noise sequence indicating only the user.

To this end, the pseudo-noise sequence generator 12 receives the user key to generate a pseudo-noise sequence Un corresponding to the user key, and provides it to the watermark generator 15 (S214). At this time, the pseudo-noise sequence Un generated by the pseudo-noise sequence generator 12 represents values of U ₁ , U ₂ , U ₃ , ..., U _N according to the user key.

The watermark generator 15 receives the similar noise sequence Un and generates a watermark Wn corresponding to the similar noise sequence Un, wherein the watermark Wn is the pseudo noise sequence Un. In accordance with, W ₁ , W ₂ , W ₃ , ...., W _N (S215).

The watermark generator 15 adjusts the intensity of the watermark Wn to have an intensity corresponding to the intensity (amplitude value) of the point to be inserted with respect to the generated watermark so as to be less affected by the background noise noise. The intensity at this time is determined by the value of f (H, Un). Here, H may be referred to as the intensity of the peripheral region of the point where the watermark is inserted, and the intensity of the peripheral region of I _n (X, Y). There are many ways to find H, but the simplest and easiest way is to find the strength of the set area surrounding the insertion point, average these values, average the mean value x Un, and return the result as H. It is.

Therefore, in Fig. 3, the side closer to the low frequency region has a smaller noise, so the intensity of the watermark is increased, and the side closer to the high frequency region has a larger noise, so the intensity of the watermark is increased. But for a simple implementation, Wn = Un can be.

The watermark Wn generated by the watermark generator 15 is input to the watermark embedding unit 16, and the embedding information determined by the embedding area determiner 14 is input to the watermark embedding unit 16. . Therefore, as described above, when the positions of the insertion points are determined by the insertion area determining unit 14, and the watermark Wn to be inserted by the watermark generating unit 15 is generated, the watermark embedding unit 16 Inserts the watermark Wn into the insertion points I _n (S216).

Herein, in the process of inserting Wn into I _n , the watermark inserting unit 16 performs I _n (X, Y) + W _n when W _n is positive, and rotates with I _n (X, Y) when W _n is negative. Perform pair I _n (X ', Y')-Wn. This makes it possible to add a watermark that always has a positive value.

The reason for the above is as follows: | I _n (X, Y) | The values are all zero or more, but W _n has both positive and negative values. Thus, at some point | I _n (X, Y) | When + W _n (X, Y) <0, the signal is distorted. To solve this problem, when adding a watermark to | I _n (X, Y) |, the watermark is added using (X, Y) and a point rotated by a specific angle in pairs.

For example, if 90 degrees are paired, (X, Y) and (Y, -X) are paired. In order to obtain a watermark (W _n) signal at a later detection, | I _n (X, Y) | -| I _n (Y, -X) |

As described above, when the watermark is inserted, the watermark embedding image generating unit 17 combines | I (X, Y) | with the watermark embedded and PI (X, Y) into the digital signal WI in the FFT region. (S217), 2D IFFT (Inverse FFT), (S218), and convert to a spatial domain, and then remove the zero inserted during zero padding to remove the watermark embedded digital image wi (x , y) is completed (S219).

Hereinafter, a watermark detection apparatus according to an embodiment of the present invention will be described with reference to FIGS. 4 and 5.

4 is a block diagram of a watermark detection apparatus according to an embodiment of the present invention, Figure 5 is a view showing a watermark detection process according to an embodiment of the present invention.

As shown in FIG. 4, the watermark detection apparatus according to the embodiment of the present invention includes a Fourier transform unit 21, a pseudo-noise sequence generator 22, a component separator 23, and an extraction region determiner 24. And a watermark generator 25, a white filter 26, a correlator 27, and a watermark discriminator 28.

Since the Fourier transform unit 21, the pseudo noise sequence generator 22, the component separator 23, and the watermark generator 25 perform the same operations as those described with reference to FIGS. Do not explain.

The extraction area determiner 24 generates J _m by selecting a point corresponding to the user's key on the straight line passing through the origin of the 2D FFT amplitude region as the point where the watermark is inserted, and selecting the insertion point thereafter. Follow the control of the discriminating unit 28.

The white filter 26 creates an autocorrelation matrix using arbitrary straight lines passing through the origin in the frequency domain of the amplitude component in order to minimize the interference effect of the original signal I _n , and obtains the whitening filtering matrix from the autocorrelation matrix. The watermark generator 25 provides the correlator 27 with the watermark Wn.

Here, a process of obtaining the whitening filtering matrix in the white filter 26 will be described.

The white filter 26 applies | I _n (X,) to each of the straight lines of any straight line M _a (a = 1, ..... A) passing through the origin among the amplitude components provided by the component separator 23. As in the case of selecting Y) |, a point is selected using the user's key to estimate that the watermark is inserted, and a vector B _a of (K, 1) dimensions consisting of K points is created. At this time, there are A vectors B. As an example, we can think of A = 180 straight lines from 1 degree to 180 degrees.

Then, the white filter 26 obtains the autocorrelation matrix R in the (K, K) dimension through R (autocorrelation matrix) = ΣB _a B _a ^T. Then, two new eigenvector matrices (V) and eigenvalue matrices (Λ) are obtained through eigenvalue decomposition of the autocorrelation matrix. Eigenvalue decomposition is as shown in Equation 3 below.

R {v} _ {i} = {lambda} _ {i} {v} _ {i}, & i = 1, 2, ...., K

R is an autocorrelation matrix, v _i and λ _i are eigenvectors and eigenvalues, respectively.

Here, the eigenvector matrix V is as shown in Equation 4 below.

V = [{v} _ {1}, ..., {v} _ {K}]

The eigenvalue matrix Λ is a diagonal matrix as shown in Equation 5 below.

 LAMBDA = diag ({lambda} _ {1}, ..., {lambda} _ {K})

The eigenvalue decomposition of the autocorrelation matrix is given by Equation 6 below.

R = V LAMBDA {V} ^ {T}

From the above Equations 3 to 6, the whitening filtering matrix Q as shown in Equation 7 is obtained.

Q = V (LAMBDA + gamma I) ^ {-1} {V} ^ {T}

Here, γ prevents an error due to singular value detection as a toughness coefficient. I is a unit matrix.

The white filter 26 outputs the whitening filtering matrix Q generated as described above and the watermark Wn generated by the watermark generator 25 to the correlator 27.

The correlator 27 receives the watermark Wn, the whitening filtering matrix Q, and the J _m selected by the extraction region determiner 24 to perform autocorrelation to be inserted into j (x, y). The watermark will be extracted.

Hereinafter, the operation of the watermark detection apparatus according to the embodiment of the present invention will be described with reference to FIG. 5.

First, j (x, y), which is a digital image input to the detection apparatus of the present invention, has a watermark inserted therein, and is horizontally or vertically deformed, rotated or translated by an operator or not.

The digital image j (x, Y) is zero-padded in a forward direction of a size of 2T × 2T by the Fourier transform unit 21 so as to be equal to the zero patting size at the time of embedding the watermark (S510), and the 2D FFT is Fourier transformed. It changes to J (X, Y) (S511).

J (X, Y), which is a 2D FFT-converted digital image, is input to the component separator 23, and the component separator 23 separates the amplitude component and the phase component (S512).                     

 The present invention is characterized in that the watermark is embedded in the amplitude component. Accordingly, the watermark embedding point in the amplitude region is selected by outputting the amplitude component in the frequency domain in which the watermark is inserted to the extraction region determiner 24. To be.

The extraction region determiner 24 receives the amplitude region signal of J (X, Y) to determine where the watermark is inserted in the amplitude region. However, the extraction area determiner may not know the information on whether the input j (x, y) is rotated, moved in parallel, or deformed in scale. Therefore, using the information in which the watermark is inserted as described with reference to Figs. 1 to 3, the point where the watermark is likely to exist is determined first. At this time, the guessing uses the user key used at the time of embedding the watermark.

Thus, the point estimated by the extraction area determiner 24 will be on a number of straight lines among all straight lines passing through the origin. Since a plurality of straight lines in which the watermark is inserted will have the same number (K) of watermarks inserted therein, the number of the plurality of straight lines in which the watermark is inserted is M and the number of watermarks inserted in one straight line If there are K pieces, the insertion point determined by the extraction area determining unit 24 is K × M pieces.

Referring to the operation of the extraction area determination unit 24, the extraction area determination unit 24 passes the origin through the user key in the same manner as the | I _n (X, Y) | From the points on each straight line (L ₁ , ....., L _M ), the point at which the watermark is inserted is sampled by sampling in the same way as the insertion on the straight line. In this case, the points obtained through the straight lines represent a sequence value, and when the sequence at this time is J _n , J _n is obtained through the sequence J _n = W _n + I _n and is provided to the correlator 27 (S213). .

At this time, the offset is adjusted so that the average becomes 0 in constructing the J _n sequence.

Meanwhile, the similar noise sequence generator 22 receives the user key to generate a similar noise sequence Un corresponding to the user key, and provides the same to the watermark generator 15 (S514). At this time, the pseudo-noise sequence Un generated by the pseudo-noise sequence generator 12 represents values of U ₁ , U ₂ , U ₃ , ..., U _N according to the user key.

The watermark generator 25 receives the similar noise sequence Un and generates a watermark Wn corresponding to the similar noise sequence Un and provides the watermark Wn to the white filter 26. ) Follows W ₁ , W ₂ , W ₃ , ...., W _N because they are based on the pseudo-noise sequence Un (S515).

The white filter 26 obtains the whitening filtering matrix Q for the amplitude region of the digital signal into which the rotation, horizontal, vertical size variation, and parallel shifted watermark are inserted, and then the whitening filtering matrix Q and the user key. The watermark (Wn) corresponding to is outputted to the correlator 27 (S516).

The J _n has K points on _M straight lines (L ₁ ,..., L _M ) passing through the origin, so that the total length is M × K = N, and the correlator 27 detects this. , J _n and W _n are divided into M sequences of length K, taking the autocorrelation by piece wise (S517), and adding the M autocorrelation values to the total correlation value (S518). .

That is, the correlator 27 divides the J _n , W _n , and I _n sequences into M sequences of length K for piecewise, and constitutes M (K, 1) -dimensional vectors B _a , where { J _n } → { J ₁ , ...., J _M }, {W _n } → { W ₁ , ...., W _M }, {I _n } → { I ₁ , ...., I _M }. Here, { J ₁ , ...., J _M }, { W ₁ , ...., W _M }, { I ₁ , ...., I _M } represent a vector or a matrix.

This operation is a piece-wise correlation operation. The piece-wise correlation operation can be expressed as shown in Equation 8, and the piece-wise correlation value C _m is calculated by the piece-wise correlation operation.

The correlator 27 provides the water mark discrimination unit 28 with C _m obtained through Equation 8 above.

The watermark discrimination unit 28 obtains a discrimination value C by averaging the C _m values, and compares the obtained discrimination value C with a set threshold value and extracts the watermark when the discrimination value C is smaller than the threshold value. If the determination value C is larger than the threshold, it is determined that the watermark is detected. If it is determined that the watermark is detected, the watermark is inserted at a point corresponding to J _n selected by the extraction area determiner 24, and the watermark is extracted by extracting the watermark.

Here, as described above, the detection device of the present invention is said that it is not known how j (x, y) is geometrically modified, thereby assuming that there is no geometrical deformation detection of the present invention Operation was performed. Therefore, if it is determined by the watermark discriminating unit 28 that no watermark has been detected, this means that j (x, y) is geometrically deformed.

Therefore, if it is determined that the watermark detection unit 28 has not detected the watermark, the watermark determination unit 24 causes the extraction area determination unit 24 to change the watermark insertion point J _m so that the watermark can be detected. Allow the filtering matrix (Q) to change.

In accordance with the change signal of the watermark determination unit 28 as described above, the extraction area determination unit 24 performs a process of obtaining the J _n sequence corresponding to the rotation and the horizontal and vertical reduction / enlargement.

That is, when the extraction area determiner 24 receives a signal from the watermark discrimination unit 28 that the watermark has not been detected, the extraction region determiner 24 determines that the watermark has been inserted, and is rotated by 1 ° from the first selection straight lines selected. Is selected, and K points on the selected straight lines are generated using the user key, and a new J _n is provided to the correlator 27. The white filter 26 obtains a new whitening filtering matrix corresponding to the new K point and provides it to the correlator 27 so that the watermark discriminating unit 28 can newly determine whether the watermark is detected.

Herein, the present invention is not limited to a straight line rotated by 1 ° with respect to the previous selected straight line to generate a new J _n , but may be a straight line rotated by 2 °, 3 °, or 10 °.

The extraction area determination unit 24 of the present invention performs a process of the extraction area determination unit 14 selecting a straight line rotated by 1 ° with respect to the previous straight line when the watermark determination unit 28 fails to detect the watermark. Repeat.

In addition to the operation corresponding to the rotation as described above, the extraction area determination unit 24 moves the K points on the previously selected straight line to correspond to the scale change by a predetermined percentage by a predetermined percentage with respect to the total length on the straight line. Create a new J _n and provide it to the correlator 27.

The reason for the above is that even when the geometric deformation is performed, the straight line passing through the origin is rotated about the origin when rotating, and only the length of the straight line is reduced or enlarged when the horizontal / vertical size is deformed, and the parallel movement is irrelevant. to be.

Therefore, J _n sequence is obtained while moving from 1 ° to 180 ° by unit angle, that is, 1 °, and the search is 1/20 to 2 times 0.50, 0.51, 0.52, ..., 1.98, 1.99, 2.00 If it is performed in units of%, the detection process of about 150 times 180 degrees = 27000 times is repeated.

If the watermark determination unit 28 determines that the determination value C is larger than the threshold while the detection process is repeated, it is determined that the watermark is extracted.

In the above-described detection apparatus of the present invention, the use of a white filter has been described as an example, but it will be readily apparent to those skilled in the art that the object of the present invention can be achieved without using a white filter.                     

In one example, the present invention may be accomplished in other embodiments as follows.

The watermark detection apparatus according to another embodiment of the present invention is characterized in that it uses a matching filter without using a white filter, and the overall operation is described with reference to FIGS. 4 and 5 except that the white filter is not used. Same as one description.

Specifically, in the watermark detection apparatus according to another embodiment of the present invention, a portion different from the embodiment described with reference to FIGS. 4 and 5 does not use a white filter, and performs a function of the matching filter in the correlator. At this time, the correlator 27 receives the watermark Wn and J _n , and performs a piecewise correlation operation as shown in Equation 9 below.

Where I is the identity matrix.

Meanwhile, the watermark detection apparatus according to another embodiment of the present invention uses a white filter and a matched filter. If only the white filter is used, the amount of data operation for detecting the watermark increases, and thus, the processing speed may be slowed. If only the matched filter is used, the whitening noise may not be removed.

Accordingly, the watermark detection apparatus according to another embodiment of the present invention compares C, which is an average value of the correlation value C _m calculated by the correlator, with respect to the set two threshold values, the first threshold value and the second threshold value. It is to decide when to use the white filter. The use time determination is performed by the watermark determination unit. The first threshold is greater than the second threshold.

In more detail, another embodiment of the present invention performs a process of initially detecting a watermark using only a matched filter, and compares the average value C obtained from the output of the correlator with a second threshold in a watermark discriminator. If the second threshold value is greater than or equal to the second threshold value, the watermark detection probability is determined to use a white filter. If the average value C obtained while using the white filter is greater than or equal to the first threshold value, the watermark is determined to be detected.

The technical spirit of the present invention has been described above with reference to the accompanying drawings, but this is by way of example only and not intended to limit the present invention. In addition, it is obvious that any person skilled in the art can make various modifications and imitations without departing from the scope of the technical idea of the present invention.

When the watermark is inserted and detected according to the present invention, the watermark can be detected regardless of this even when the RST transformation is applied to the image.

Claims (35)

A pseudo noise sequence generator for receiving a user key from the outside and generating a pseudo noise sequence corresponding to the user key; A Fourier transform unit that zero-pads a digital image and performs a 2D Fast Fourier Transform (2D FFT); A component separator configured to divide an amplitude component and a phase component in the two-dimensional fast Fourier transform digital image; An insertion area determination unit which selects a straight line passing through the origin point using the user key in the amplitude area obtained by the component separation unit, and determines a point corresponding to the user key among the points on the selected straight line as the watermark insertion point. ; A watermark generator for generating a watermark using a pseudonoise sequence generated by the pseudonoise sequence generator; A watermark inserting unit for inserting the watermark generated by the watermark generating unit so as to match the point determined by the embedding area determining unit; And A watermark embedding image generating unit for generating a digital image embedding a watermark through two-dimensional inverse fast Fourier transform and zero removing of the amplitude region into which the watermark is inserted and the phase component by the watermark inserting unit; Watermark embedding apparatus comprising a. The method of claim 1, The insertion area determiner, In the amplitude component located on a straight line passing through the origin, the watermark insertion point is determined for an amplitude component whose radial distance from the origin is between R1 and R2, The radial distance R1 is a boundary between the predetermined low frequency region and the intermediate frequency region, The radial distance R2 is a boundary line between the predetermined intermediate frequency region and the high frequency region. Watermark insertion apparatus, characterized in that. The method of claim 1, The insertion area determiner, In order to reduce the quantization error occurring in rotation and scaling, the watermark embedding apparatus, characterized in that the first point and the points surrounding the first point in common to determine the watermark embedding point. delete The method of claim 1, The watermark inserting unit, A watermark embedding apparatus, characterized in that the length of the watermark to be inserted is divided into M subsets and inserted on different straight lines. The method of claim 1, The watermark inserting unit, When inserting a positive and negative watermark, watermark insertion is characterized by always inserting a positive watermark and a negative watermark as positive numbers at two different points having a correlation with each other, considering that the background noise is always positive. Device. A pseudo noise sequence generator for receiving a user key from the outside and generating a pseudo noise sequence corresponding to the user key; A Fourier transform unit which zero-pads the digital image to perform two-dimensional fast Fourier transform; A component separator configured to divide an amplitude component and a phase component in the two-dimensional fast Fourier transform digital image; An extraction region determination unit which selects a straight line passing through the origin point using a user key in the amplitude region obtained by the component separation unit, and determines points corresponding to the user key among points located on the selected straight line as watermark extraction points; A watermark generator for generating a watermark using a pseudonoise sequence generated by the pseudonoise sequence generator; A correlator for calculating a correlation value using a random number sequence (J _n) of the watermark extraction point provided by the sequence of the watermark (W _n) and the extraction region determination section generated by the watermark generating unit; And And a watermark discriminating unit for determining whether a watermark is detected using a correlation value of the correlator and a set first threshold value. delete The method of claim 7, wherein The sequence of watermarks generated by the watermark generation unit (W _n ) and a user key are received, an autocorrelation matrix of a region where a watermark is inserted is obtained, and a whitening filtering matrix (Q) is obtained from the autocorrelation matrix. obtain, watermark detection apparatus for the whitening filter matrix (Q) and the sequence (W _n) of the watermark, characterized in that it further comprises a white filter provided to the correlator. delete delete The method of claim 9, The white filter, Obtain an autocorrelation matrix from points located on a straight line passing through the origin, with the radial distance from the origin between R1 and R2, The radial distance R1 is a boundary between the predetermined low frequency region and the intermediate frequency region, And the radial distance R2 is a boundary line between a predetermined intermediate frequency region and a high frequency region. The method of claim 9, The watermark generation unit, Generates a watermark during initial driving and provides it directly to the correlator without passing through the white filter. When a white filter driving control signal is input from the watermark discriminating unit, the generated watermark is provided to the white filter. The white filter is not driven when the watermark is not provided, but is driven according to a control signal of the watermark discrimination unit to receive a watermark from the watermark generator and to receive the watermark sequence W _n and the white color. And providing a sum filtering matrix to the correlator. delete The method of claim 7, wherein The watermark determination unit, When inserting a positive and negative watermark, the background noise is always positive, using a feature that is always inserted as a positive number at two different points having a positive correlation with a positive watermark and a negative watermark. The watermark detection apparatus, characterized in that it is determined whether the watermark is detected by the difference between the values of the two different points. delete The method of claim 7, wherein The extraction region determiner, In order to reduce the quantization error occurring in rotation and scaling, the watermark detection apparatus, characterized in that the first point and the points surrounding the first point in common to determine the watermark extraction point. delete A first step of generating a similar noise sequence using a user key input from the outside; A second step of zero-padding the original signal to be inserted into the watermark to be input from the outside and a two-dimensional fast Fourier transform key; A third step of separating components into amplitude components and phase components in the two-dimensional fast Fourier transform digital image; Selecting a straight line including amplitude components passing through the origin in the amplitude region obtained by the component separating unit, and determining a point corresponding to the user key among points located on the selected straight line as a watermark insertion point; A fifth step of generating a watermark corresponding to the similar noise sequence and inserting the watermark at the insertion point determined in the fourth step; And And a sixth step of making the amplitude region into which the watermark is inserted into a zero-deleted digital image. The method of claim 19, The fourth step, A watermark insertion point is determined at a point located on a straight line passing through the origin, with a radius distance from the origin between R1 and R2. The radial distance R1 is a boundary between the predetermined low frequency region and the intermediate frequency region, The radial distance R2 is a boundary line between the predetermined intermediate frequency region and the high frequency region. Watermark embedding method, characterized in that. The method of claim 20, The fourth step, In order to reduce the quantization error occurring in rotation and scaling, the watermark embedding method, characterized in that the first point and the points surrounding the first point in common to determine the watermark embedding point. delete The method of claim 19, The fifth step, And dividing the length of the watermark to be inserted into M subsets and inserting them on different straight lines. The method of claim 19, The fifth step, When inserting a positive and negative watermark, watermark insertion is characterized by always inserting a positive watermark and a negative watermark as positive numbers at two different points having a correlation with each other, considering that the background noise is always positive. Way. A first step of receiving a user key from an external source and generating a similar noise sequence corresponding to the user key; A second step of zero padding the digital image into which the watermark input from the outside is inserted and converting the two-dimensional fast Fourier transform; Dividing the two-dimensional fast Fourier transform digital image into an amplitude component and a phase component; Selecting a straight line including amplitude components passing through the origin in the amplitude region obtained by the component separation unit, and determining points J _n corresponding to the user's key among the points on the selected straight line as the watermark extraction point; A fourth step; A fifth step of calculating a correlation value using a random number sequence (J _n) of the watermark extraction point provided by the sequence of the watermark (W _n) and the extraction region determination section generated by the watermark generating unit; And And a sixth step of determining whether a watermark is detected using the correlation value and the set first threshold value. delete The method of claim 25, The autocorrelation matrix of the region in which the watermark is inserted is obtained by receiving the sequence W _n of the watermark and the user key, and the whitening filtering matrix Q is obtained from the autocorrelation matrix, and the whitening filtering matrix Q is obtained. ) And a seventh step of providing the sequence of watermarks (W _n ) to the fifth step. delete delete The method of claim 27, The seventh step, The autocorrelation matrix is obtained from the points located on a straight line passing through the origin, and the radial distance from the origin lies between R1 and R2. The radial distance R1 is a boundary between the predetermined low frequency region and the intermediate frequency region, The radial distance R2 is a boundary line between the predetermined intermediate frequency region and the high frequency region. Watermark detection method, characterized in that. The method of claim 25, The sixth step, Comparing the average value of the correlation value and the first threshold value, it is determined that the watermark is detected when the average value is larger than the first threshold value, and that the watermark is not detected when the average value is less than the first threshold value. Watermark detection method, characterized in that. delete delete delete delete

Images