CN101364300A - Digital watermarking method based on gray theory - Google Patents

Abstract

The invention discloses a digital watermarking method based on a grey theory, which mainly solves the contradiction problem between the embedded information quantity and the watermarking transparency and has the technical proposal that the grey theory and the embedding as well as the extraction of the watermarking are combined together organically; an original host image is partitioned and the grey relational grade is calculated during the image embedding process so as to determine the embedding area and reduce the overall influence to the host image caused by the watermarking, and the watermarking embedding is completed within the discrete wavelet transform; during the watermarking extraction process, the watermarking image is treated through zero setting reservation, the information quantity of the embedding watermarking is reduced, the watermarking transparency is increased, the extracted watermarking image is processed by utilizing a grey forecasting model, and the watermarking information is restored. The digital watermarking method has the advantages of good transparency and robustness, and can be used for protecting the copyright of digital works.

Description

Digital watermark method based on gray theory

Technical field

The invention belongs to the numerical information field, relate to numeric security, particularly relate to the digital watermark method of protection copyright safety, can be used for copyright protection copyright.

Background technology

In today of rapid development of information technology, information security is more and more noticeable, digital watermark technology big effect of play more and more aspect the protection information security.Because the quantity of information that embeds has become a pair of contradiction with the transparency of watermark, thereby how can embed the information of enough expression copyrights, guarantee that again the transparency of watermark becomes the technical matters that presses for solution simultaneously.In recent years, the application of relevant gray theory on water mark method has certain development, gray theory comprises grey relation analysis method and gray prediction method, grey relation analysis method has remedied and has adopted mathematical statistics method to make the shortcoming that systematic analysis causes, it to sample size what and have irregular all suitable equally, and calculated amount is little, and is very convenient, more the situation that quantized result and The qualitative analysis are not inconsistent can not occur.Gray prediction is an important component part of gray system theory.By this new curve fitting technique of accumulative being incorporated among the parameter estimation of gray model, avoided complicated matrix operation, reduced operand, precision is high and do not change many character of model simultaneously, for practical application certain meaning is arranged.

Relevant digital watermark method or the method relevant with watermark that relates at present gray theory both at home and abroad mainly contains:

(1). the method for utilizing UGM model extraction watermark of propositions such as Li Feng, Li Feng, grey prediction is theoretical and application with the ash assessment, Central China University of Science and Technology's doctorate paper, 2002.10, this method mainly is to utilize the UGM model to recover watermark information.

(2). the judge that is applicable to the evaluation watermark transparency of propositions such as Ma Miao, Ma Miao, fourth power, Hao Chongyang, the ash of the transparency of digital watermarking is passed judgment on, computer engineering and application, 2003.22, this method is estimated the transparency of watermark with grey relevance theory.

(3). the grey relevance theory of the usefulness of propositions such as Tian Hongpeng is estimated the transparent algorithm of watermark, Tian Hongpeng, Ma Miao, watermark transparency evaluation algorithms based on the grey correlation theory, computer engineering and application, 2006.23 this method comes the transparency of watermark is made objective quantitative evaluation with grey relevance theory.

The greatest drawback that above-mentioned these existing methods exist is: only used gray prediction to recover watermarking images, perhaps only pass judgment on the watermark transparency with grey association knowledge, do not have grey associated application in the embedding and extraction of watermark, can't solve the transparency of watermark and the contradiction of embedding quantity of information, influence the effect of digital watermarking reference numbers works copyright.

The content of invention

The objective of the invention is to overcome the deficiency of above-mentioned prior art, a kind of digital watermark method based on gray theory is proposed, the knowledge of grey correlation analysis and grey prediction and watermark embedded and extract organically combine, to guarantee that watermark embeds the copyright information amount of sufficient intensity under the condition of the transparency, improve the robustness of watermark.

The technical scheme that realizes the object of the invention is: utilize grey relevance theory to choose and embed the district, watermarking images is carried out reservation process, realized that the embedding of watermark is extracted, and recovered watermarking images with the grey prediction model.Detailed process is as follows:

One, digital watermarking embeds

1 carries out piecemeal according to the size of watermark to original host image I, and calculates the grey degree of association sum of each piece;

2. the piece of choosing grey degree of association minimum from the image of institute's piecemeal is as embedded block A, and this piece is carried out wavelet transform, obtains high-frequency sub-band HH, medium-high frequency subband HL, LH, low frequency sub-band LL;

3. watermarking images W is divided into several one-dimension array, array length determined by the size of watermarking images, and back two information in each one-dimension array are changed to 0, and reconstructuring water-mark image T carries out scramble to the watermarking images T of reconstruct and handles, and obtains watermark W1 behind the scramble;

4. watermark W1 behind the scramble is utilized mixed formulation, $\left[\begin{array}{c}F1\\ F2\end{array}\right]\left[\begin{array}{cc}a1& 1-a1\\ a2& 1-a2\end{array}\right]\left[\begin{array}{c}\mathrm{<figure-callout\; id="1"\; label="SS\; W"\; filenames="A200810018338E00162.png"\; state="\{\{state\}\}">SS</figure-callout>}& \\ W1\end{array}\right]$ Be embedded among the low frequency sub-band LL of host image, in the formula, SS is a to be embedded low frequency sub-band LL part after process one deck wavelet decomposition, F1 is the moisture mark band behind the embed watermark, F2 is for keeping the key of recovering information, and a1 embeds coefficient for mixing, and a2 is that key generates coefficient;

5. will contain the low frequency sub-band LL and the high-frequency sub-band HH of watermark information, medium-high frequency subband HL, LH carry out the discrete wavelet inverse transformation together, obtain containing the image block J of watermark;

6. replace described embedded block A with the image block J that contains watermark, and with other not the image block of embed watermark splice, reconstruct and contain watermarking images P.

Two, digital watermarking is extracted

1. the image P that will contain watermark is divided into four: (i (1,1), i (1,2), i (2,1) and i (2,2)), find the image block J that contains watermark;

2. will contain watermarking images piece J and carry out wavelet transform, and obtain this and contain watermarking images piece high-frequency sub-band HH, medium-high frequency subband HL, LH, low frequency sub-band LL, and this is contained the low frequency sub-band LL SS symbolic representation of watermarking images piece;

3. utilize the watermark extracting formula $W$ $2=\frac{F2-a2\&times;\mathrm{SS}}{1-a2},$ Extract the watermarking images behind the scramble, and, obtain the watermarking images Q that disorderly handles through being inverted the watermarking images W2 of this extraction unrest that is inverted;

4. utilize grey forecasting model to handling, recover watermarking images WW through the watermarking images Q that disorderly handles that is inverted.

The present invention and existing technology relatively have following advantage:

1, the present invention calculates the grey relational grade of each piece owing to adopt the host image piecemeal, and the image block of seeking the correlation degree minimum is as embedded block, so the modification of this embedded block to the minimum that influences of entire image, can guarantee the transparency of watermark;

2, the present invention has reduced the watermark information amount that embeds owing to adopt the reservation process of watermark information being carried out last two positions zero, has increased the embedment strength that keeps information, has effectively guaranteed the transparency and the robustness of watermark;

3, the present invention is owing in the process of extracting, utilize grey forecasting model to handle the watermark of extracting, thereby the watermark information that recovers is complete and accurate.

4, the present invention is chosen in the embedding that watermark is finished in the wavelet transform territory, so have the advantage that the watermark of discrete wavelet territory is had.

Description of drawings

Fig. 1 is a digital watermark embed process block diagram of the present invention;

Fig. 2 is a host image grey correlation analysis piecemeal synoptic diagram of the present invention;

Fig. 3 is the wavelet decomposition synoptic diagram;

Fig. 4 is a watermarking images reservation process piecemeal synoptic diagram of the present invention;

Fig. 5 is the block diagram of choosing of host image embedded block of the present invention;

Fig. 6 is that the present invention embeds the preceding processing block diagram to watermark of host image;

Fig. 7 is a digital watermarking leaching process block diagram of the present invention;

Fig. 8 is that watermarking images grey prediction of the present invention recovers to handle block diagram;

Fig. 9 is that the attack front and back of emulation experiment of the present invention contain the watermarking images comparison diagram;

Figure 10 is watermarking images and the original watermark image comparison diagram that emulation experiment of the present invention is extracted and recovery is come out.

Embodiment

1. basic theory introduction

1.1 ash is related

The basic thought of grey correlation analysis is to judge according to the similarity degree of sequence curve geometric configuration whether its contact is tight, and curve is approaching more, and the degree of association between corresponding sequence is just big more, otherwise then more little.

Basic thought according to grey relevance theory, the pixel value of image object as a comparison, correlation degree between the calculating pixel, a certain block of pixels and surrounding associated degree are more little, and then it being made an amendment to compare to is modified in visually more difficult being found to the bigger block of pixels equal extent of correlation degree.The related computation process of ash is described below.

The system behavior sequence of setting up departments is:

X ₀＝(x ₀(1)，x ₀(2)，…，x ₀(n))

X ₁＝(x ₁(1)，x ₁(2)，…，x ₁(n))

..................

X _i＝(x _i(1)，x _i(2)，…，x _i(n))

..................

X _m＝(x _m(1)，x _m(2)，…，x _m(n))

For ζ ∈ (0,1), order

$\mathrm{\&gamma;}(x_0\left(k\right),x_i\left(k\right))=\frac{\underset{i}{\min }\underset{k}{\min }|x_0\left(k\right)-x_i\left(k\right)|+\mathrm{\&zeta;}\underset{i}{\max }\underset{k}{\max }|x_0\left(k\right)-x_i\left(k\right)|}{|x_0\left(k\right)-x_i\left(k\right)|+\mathrm{\&zeta;}\underset{i}{\max }\underset{k}{\max }|x_0\left(k\right)-x_i\left(k\right)|}---\left(1\right)$

$\mathrm{\&gamma;}({\mathrm{<figure-callout\; id="0"\; label="X"\; filenames="A200810018338E00162.png"\; state="\{\{state\}\}">X</figure-callout>}}_0,X_i)=\frac{1}{n}\underset{k=1}{\overset{n}{\mathrm{\&Sigma;}}}\mathrm{\&gamma;}(x_0\left(k\right),x_i\left(k\right))---\left(2\right)$

Wherein ζ is called resolution ratio, grey relational grade γ (X ₀, X _i) normal brief note is γ _0i, k point correlation coefficient γ (x ₀(k), x _i(k)) brief note is γ _0i(k).

The calculation procedure of grey relational grade is as follows:

The first step: ask the initial value picture or the average picture of each sequence, order

$X_i^{\&prime;}=X_i/x_i\left(1\right)=(x_i^{\&prime;}\left(1\right),x_i^{\&prime;}\left(2\right),\&CenterDot;\&CenterDot;\&CenterDot;,x_i^{\&prime;}\left(n\right)),i=\mathrm{0,1},\&CenterDot;\&CenterDot;\&CenterDot;,m$

Second step: ask difference sequence, note

${\mathrm{\&Delta;}}_i\left(k\right)=|x_0^{\&prime;}\left(k\right)-x_i^{\&prime;}\left(k\right)|$

Δ _i＝(Δ _i(1)，Δ _i(2)，…，Δ _i(n))

i＝0，1，2，…，m

The 3rd step: ask maximum difference in the two poles of the earth and lowest difference, note

$\begin{array}{cc}M=\underset{i}{\max }\underset{k}{\max }{\mathrm{\&Delta;}}_i\left(k\right),& m=\underset{i}{\min }\underset{k}{\min }{\mathrm{\&Delta;}}_i\left(k\right)\end{array}$

The 4th cloth: ask correlation coefficient

${\mathrm{\&gamma;}}_{0i}\left(k\right)=\frac{m+\mathrm{\&zeta;M}}{{\mathrm{\&Delta;}}_i\left(k\right)+\mathrm{\&zeta;}M},\mathrm{\&zeta;}\&Element;\left(\mathrm{0,1}\right)\begin{array}{}\end{array}$

k＝1，2，…，n；i＝1，2，…，m

The 5th step: compute associations degree

Can realize calculating with above step to the image ash degree of association.

1.2 grey prediction

Prediction is exactly to go supposition to understand future by means of the discussion in past.The system development rule is found, grasped to gray prediction by the processing of raw data and the foundation of gray model, and the to-be of system is made the scientific quantitative prediction.Based on the grey method of grey modeling theory, according to the feature of its forecasting problem, can be divided into five kinds of fundamental types, i.e. ordered series of numbers prediction, catastrophe prediction, catastrophe prediction in season, topology prediction and system synthesis prediction.This Forecasting Methodology of five types all is an important and common forecasting method in the regionl development research.The pixel of image as a matrix, according to the requirement of watermark, is selected the ordered series of numbers prediction.

Use gray model GM (1,1) intension type to carry out the ordered series of numbers prediction.X wherein ⁽¹⁾, Z ⁽¹⁾, a, b, C, D, E, F definition specific as follows:

Make that original series is x ⁽⁰⁾(k), k=1,2 ... n.

x ⁽⁰⁾(k)+a·x ⁽¹⁾(k)＝b (3)

x ⁽¹⁾＝(x ⁽¹⁾(t ₁)，x ⁽¹⁾(t ₂)，…x ⁽¹⁾(t _n))， $x^{\left(1\right)}\left(t_k\right)=\underset{m=1}{\overset{k}{\mathrm{\&Sigma;}}}{x}^{\left(0\right)}\left(t_m\right);$ $x^{\left(1\right)}\left(t_k\right)=\underset{m=1}{\overset{k}{\mathrm{\&Sigma;}}}{x}^{\left(0\right)}\left(t_m\right);$ Or x ⁽¹) (t _k)=x ⁽¹⁾(t _K-1)+x ⁽⁰⁾(t _k)

Z ⁽¹⁾Be X ⁽¹⁾The MEAN sequence

z ⁽¹⁾＝(z ⁽¹⁾(t ₁)，z ⁽¹⁾(t ₂)，…z ⁽¹⁾(t _n))， (4)

z ⁽¹⁾(t _k)＝0.5x ⁽¹⁾(t _k)+0.5x ⁽¹⁾(t _k-1) (5)

$C=\underset{k=2}{\overset{n}{\mathrm{\&Sigma;}}}{z}^{\left(1\right)}\left(t_k\right)---\left(6\right)$

$D=\underset{k=2}{\overset{n}{\mathrm{\&Sigma;}}}{x}^{\left(0\right)}\left(t_k\right)---\left(7\right)$

$E=\underset{k=2}{\overset{n}{\mathrm{\&Sigma;}}}{z}^{\left(1\right)}\left(t_k\right)x^{\left(0\right)}\left(t_k\right)---\left(8\right)$

$F=\underset{k=2}{\overset{n}{\mathrm{\&Sigma;}}}{z}^{\left(1\right)}{\left(t_k\right)}^2---\left(9\right)$

$a=\frac{\mathrm{CD}-(n-1)E}{(n-1)F-C^2}---\left(10\right)$

$b=\frac{\mathrm{DF}-\mathrm{CE}}{(n-1)F-C^2}---\left(11\right)$

And the expression formula of intension type is:

${\hat{X}}^{\left(0\right)}\left(k\right)=u^{k-2}\&times;r,---\left(12\right)$

Wherein,

$\begin{array}{cc}u=\left(\frac{1-0.5\&times;a}{1+0.5\&times;a}\right),& v=\left(\frac{b-a\&times;x^{\left(0\right)}\left(1\right)}{1+0.5\&times;a}\right)\end{array}.$

The present invention utilizes this GM (1,1) intension pattern type just, and the watermarking images after the reservation process is predicted recovery.

2. based on the specific implementation of the digital watermarking of gray theory

2.1 related symbol explanation:

I: original host image

W: original watermark image

A: original host image embedded block

SS: embedded block is carried out low frequency LL part after one deck wavelet decomposition

T: original watermark image is carried out the watermarking images that reservation process remains later on

W1: the image that the later watermarking images scramble of reservation process obtains later on

H: degree of association summation

J: original host image embedded block inverse transformation image of returning again after DWT conversion embed watermark

P: contain watermarking images

W2: the scramble watermark that extracts

Q: the watermark that is inverted and disorderly recovers

WW: the watermark that utilizes gray prediction to recover

F1: the moisture mark band behind the embed watermark

F2: the key that keeps recovering information

A1: mix embedding coefficient

A2: key generates coefficient

2.2 the embedding of digital watermarking

With reference to Fig. 1, watermark of the present invention embeds and comprises the steps:

Step 1, host image I is carried out piecemeal, calculate the grey degree of association sum of each piece.

(1a) host image I is divided into equal-sized four bulk image: I (1,1), I (1,2), I (2,1) and I (2,2);

(1b) a bulk image being divided into several sizes is 2 * 2 fritter, and gets this bulk image and remove part beyond the edge fritter as processing region, as shown in Figure 2;

(1c) with in the processing region one 2 * 2 fritter and transform into the capable vector of one dimension with its eight adjacent fritter, utilize degree of association formula: $\mathrm{\&gamma;}(X_0,X_i)=\frac{1}{n}\underset{k=1}{\overset{n}{\mathrm{\&Sigma;}}}\mathrm{\&gamma;}(x_0\left(k\right),x_i\left(k\right))$ Obtain the degree of association of the capable vector of one dimension that the capable vector of one dimension that this 2 * 2 fritter transforms and eight fritters on every side transform respectively, in the formula, X ₀Be the capable vector of one dimension that this 2 * 2 fritter transforms, X _iBe a capable vector of the one dimension that is transformed among eight fritters on every side, x ₀(k) and x _i(k) be X respectively ₀And X _iIn element, the n in the formula represents the length of the capable vector of one dimension, gets 4 here.Finish once-through operation and just calculated a degree of association, finish after the capable the calculating vectorial and degree of association of the capable vector of one dimension of eight fritter conversions on every side of one dimension of this 2 * 2 fritter conversion, then obtained eight degrees of association.Described eight degrees of association of obtaining are sued for peace, obtain the degree of association sum h (1) of the capable vector of one dimension that the capable vector of one dimension that one 2 * 2 fritter transforms and eight fritters on every side transform;

(1d) each 2 * 2 fritter in the processing region are done same processing according to step (1c), and the result that will handle adds up, obtain the degree of association summation H1 in this bulk Flame Image Process zone;

(1e) repeating step (1b) is obtained degree of association summation H2, H3 and the H4 in the Flame Image Process zone of other three bulks respectively to (1d).

Step 2, choose embedded block and carry out wavelet transform.

(2a) select that minimum among degree of association summation H1, H2, H3 and the H4 in Flame Image Process zone of each bulk bulk as embedded block A, referring to Fig. 5;

(2b) embedded block A is carried out one deck wavelet transform, obtain high-frequency sub-band HH, medium-high frequency subband HL, LH and low frequency sub-band LL, as shown in Figure 3.

Step 3, reconstructuring water-mark image and scramble are handled.

With reference to Fig. 6, the process that reconstructuring water-mark image and scramble are handled is:

(3a) determine the length of one-dimension array according to the size of watermarking images, and original watermark image W is divided into several arrays of 1 * 8: [w (1), w (2), w (3), w (4), w (5), w (6), w (7), w (8)] according to determined length;

(3b) the 7th and the 8th each array of 1 * 8 is changed to 0, only keeps the information of first six digits, the dope vector that obtains keeping [w (1), w (2), w (3), w (4), w (5), w (6), 0,0], as shown in Figure 4;

(3c) each one-dimension array that original watermark image W branches away is handled, obtained the dope vector of all reservations, with the dope vector splicing of these reservations, reconstruct the watermarking images T that remains again according to step (3b);

(3d) utilize the arnold scrambling algorithm that the watermarking images T that remains is carried out scramble and handle, obtain watermark W1 behind the scramble.

Step 4, embed watermark.

Watermark W1 behind the scramble is utilized mixed formulation, $\left[\begin{array}{c}F1\\ F2\end{array}\right]\left[\begin{array}{cc}a1& 1-a1\\ a2& 1-a2\end{array}\right]\left[\begin{array}{c}\mathrm{<figure-callout\; id="1"\; label="SS\; W"\; filenames="A200810018338E00162.png"\; state="\{\{state\}\}">SS</figure-callout>}& \\ W1\end{array}\right]$ Be embedded in the low frequency sub-band of host image,

In the formula, SS represents to be embedded through the low frequency sub-band LL part after one deck wavelet decomposition,

F1 is the moisture mark band behind the embed watermark, and F2 is the key of reservation recovering information,

A1 embeds coefficient for mixing, and the a1 value is more than or equal to 1, and it approaches 1 more, and embedment strength is just more little, promptly controls the intensity of embedding with a1, and this example is got a1=1.05,

A2 is that key generates coefficient, and this example is got a2=0.5.

Step 5, discrete wavelet inverse transformation.

The low frequency sub-band F1 and the high-frequency sub-band HH of watermark information will be contained, medium-high frequency subband HL, LH carry out the discrete wavelet inverse transformation together, obtain containing the image block J of watermark, replace embedded block A with the image block J that contains watermark, and with other not the image block of embed watermark splice, reconstruct and contain watermarking images P.

2.3 the extraction of digital watermarking

As follows with reference to Fig. 7 digital watermarking extraction step of the present invention:

Step 1, to containing the watermarking images piecemeal.

The image P that contains watermark is divided into four: (i (1,1), i (1,2), i (2,1) and i (2,2)), find the image block J that contains watermark;

Step 2, carry out wavelet transform to containing the watermarking images piece.

To contain watermarking images piece J and carry out wavelet transform, and obtain this and contain watermarking images piece high-frequency sub-band HH, medium-high frequency subband HL, LH, low frequency sub-band LL, and this is contained the low frequency sub-band LL SS symbolic representation of watermarking images piece;

The extraction of step 3, watermark and the unrest that is inverted.

Utilize the watermark extracting formula $W$ $2=\frac{F2-a2\&times;\mathrm{SS}}{1-a2},$ Extract the watermarking images behind the scramble, and, obtain the watermarking images Q that disorderly handles through being inverted the watermarking images W2 of this extraction unrest that is inverted;

Step 4, utilize gray prediction to handle to be inverted random watermarking images, recover watermarking images.

Referring to Fig. 8, the process that the present invention recovers watermarking images is:

4a) being divided into several one-dimension array of 1 * 8 through the watermarking images Q that disorderly handles that is inverted, and with gray model GM (1,1) intension type to one of them array [w (1), w (2), w (3), w (4), w (5), w (6), 0,0] carrying out gray prediction handles, dope and be changed to 0 the 7th information w (7) and the 8th information w (8), the one-dimension array that is restored [w (1), w (2), w (3), w (4), w (5), w (6), w (7), w (8)];

4b) according to step 4a) each 1 * 8 the one-dimension array that is divided into through the random watermarking images Q that handles that is inverted is handled, and all one-dimension array that recover are spliced, reconstruct the watermarking images WW of recovery.

Effect of the present invention can further specify by following emulation experiment

Simulated conditions: the software that emulation is used is MATLAB.

Simulated environment: emulation is to carry out under the environment of Windows XP.

Simulation result: referring to Fig. 9, Figure 10 and table 1.

Referring to Fig. 9, wherein, Fig. 9 (a) has shown the image that does not contain watermark, Fig. 9 (b) has shown the image that embed watermark is later, Fig. 9 (c) has shown that intensity is that 0.01 salt-pepper noise is attacked the later watermarking images that contains, Fig. 9 (d) shown quality coefficient be 5 JPEG compression attack later contain watermarking images, Fig. 9 (e) has shown that Gauss's low-pass filtering attacks the later watermarking images that contains, Fig. 9 (f) has shown that the rotations of 10 degree that turn clockwise attack the later watermarking images that contains, Fig. 9 (g) has shown that coefficient is that 0.001 Gauss adds to make an uproar and attacks the later watermarking images that contains, Fig. 9 (h) has shown that Wiener filtering attacks the later watermarking images that contains, and Fig. 9 (i) has shown that medium filtering attacks the later watermarking images that contains.From the contrast of Fig. 9 (a) and Fig. 9 (b) as can be seen, the similarity degree that contains watermarking images and original host image that embed watermark is later is very high, does not almost see any variation, illustrates that the watermark of the inventive method has the good transparency.

Referring to Figure 10, wherein Figure 10 (a) has shown original watermark image, Figure 10 (b) has shown the watermarking images that extracts and recover the watermarking images that contains that passes through attack from Fig. 9 (b), Figure 10 (c) has shown the watermarking images that extracts and recover from Fig. 9 (c), Figure 10 (d) has shown the watermarking images that extracts and recover from Fig. 9 (d), Figure 10 (e) has shown the watermarking images that extracts and recover from Fig. 9 (e), Figure 10 (f) has shown the watermarking images that extracts and recover from Fig. 9 (f), Figure 10 (g) has shown the watermarking images that extracts and recover from Fig. 9 (g), Figure 10 (h) has shown the watermarking images that extracts and recover from Fig. 9 (h), Figure 10 (i) has shown the watermarking images that extracts and recover from Fig. 9 (i).Contrast Figure 10 (a) original watermark image and Figure 10 (b) never attack contains the watermarking images that extracts and recover in the watermarking images, and extraction that watermark basically can be correct and recovery are come out as can be seen; From Figure 10 (c) to Figure 10 (i) as can be seen, process various attack watermarking images later on can be extracted basically and be recovered out, and extract and the watermarking images distortion that recovers to come out also less, visible the inventive method makes watermark have anti-attack ability preferably, promptly robustness is fine.

Data in the table 1 are that simulation contains watermarking images through after the various attack, Y-PSNR PSNR that the watermarking images that extracts and recover is compared with original watermark image and the numerical value of normalized correlation coefficient NC, wherein the degree of introducing noise is compared in the watermark of extracting and recovering of Y-PSNR PSNR value representation with original watermark, Y-PSNR PSNR value is big more, then the noise of Yin Ruing is more little, and the robustness of watermark is just good more.The watermark that normalized correlation coefficient NC value representation extracts and recovers and the similarity degree of original watermark, the upper limit of NC value is 1, and it approaches 1 more and just illustrates that the similarity degree of two width of cloth images is high more, and the recovery effects of watermark is good more, and the robustness of watermark is good more.

Table 1 experimental result data

The attack pattern of watermark process as seen from Table 1, includes: Gauss's low-pass filtering, Wiener filtering, medium filtering, rotation, convergent-divergent, shearing, JPEG are attacked, Gauss adds and makes an uproar and the spiced salt adds and makes an uproar.The value that Y-PSNR PSNR is compared in these watermarks of extracting and recovering after attacking with original watermark in process is higher, major part is all more than 16, illustrate that watermark that the present invention embeds is extracted later on through various attack and introducing is compared in the watermark that recovers to come out with original watermark noise seldom, the change of the watermarking images that extracts and recover to come out is very little, has good robustness.Simultaneously from table 1 as seen, normalized correlation coefficient NC value is all very high, great majority are all more than 0.96, it is very high to illustrate that its similarity degree is compared in watermark that watermark that the inventive method embeds extracts later on through various attack with original watermark, has shown that once more watermark of the present invention has good robustness.

To sum up, the good transparency of the present invention has guaranteed the disguise of embed watermark, does not cause degrading of carrier works.Good robustness of the present invention has guaranteed that copyright has very strong viability after the attack intentionally or unintentionally through various, effectively the information of reserved identities copyright.

Claims (5)

1. data waterprint embedded method based on gray theory comprises following process:

(1) according to the size of watermark original host image (I) is carried out piecemeal, and calculate the grey degree of association sum of each piece;

(2) piece of choosing grey degree of association minimum from the image of institute's piecemeal is as embedded block (A), and this piece is carried out wavelet transform, obtains high-frequency sub-band HH, medium-high frequency subband HL, LH, low frequency sub-band LL;

(3) watermarking images (W) is divided into several one-dimension array, array length is determined by the size of watermarking images, and back two information in each one-dimension array is changed to 0, reconstructuring water-mark image (T), the watermarking images (T) of reconstruct is carried out scramble handle, obtain watermark (W1) behind the scramble;

(4) watermark behind the scramble (W1) is utilized mixed formulation, $\left[\begin{array}{c}F1\\ F2\end{array}\right]=\left[\begin{array}{cc}a1& 1-a1\\ a2& 1-a2\end{array}\right]\left[\begin{array}{c}\mathrm{SS}\\ W1\end{array}\right]$ Be embedded among the low frequency sub-band LL of host image,

In the formula, SS is to be embedded, and F1 is the moisture mark band behind the embed watermark through the low frequency sub-band LL part after one deck wavelet decomposition, and F2 is for keeping the key of recovering information, and a1 embeds coefficient for mixing, and a2 is that key generates coefficient;

(5) will contain the low frequency sub-band LL and the high-frequency sub-band HH of watermark information, medium-high frequency subband HL, LH carry out the discrete wavelet inverse transformation together, obtain containing the image block (J) of watermark;

(6) replace described embedded block (A) with the image block (J) that contains watermark, and with other not the image block of embed watermark splice, reconstruct and contain watermarking images (P).

2. data waterprint embedded method according to claim 1, wherein step (1) is carried out according to the following procedure

(1a) host image I is divided into equal-sized four bulk image: I (1,1), I (1,2), I (2,1) and I (2,2);

(1b) a bulk image being divided into several sizes is 2 * 2 fritter, and gets this bulk image and remove part beyond the edge fritter as processing region;

(1c) with in the processing region one 2 * 2 fritter and transform into the capable vector of one dimension with its eight adjacent fritter, obtain the degree of association of the capable vector of one dimension that the capable vector of one dimension that this 2 * 2 fritter transforms and eight fritters on every side transform respectively, and, obtain the degree of association sum h (1) of the capable vector of these one dimensions to its summation;

(1d) each 2 * 2 fritter in the processing region are done same processing according to step (1c), and the result that will handle adds up, obtain the degree of association summation H1 in this bulk Flame Image Process zone;

(1e) repeating step (1b) is obtained degree of association summation H2, H3 and the H4 in the Flame Image Process zone of other three bulks respectively to (1d).

3. data waterprint embedded method according to claim 1, wherein step (3) is carried out according to the following procedure:

(3a) determine the length of one-dimension array according to the size of watermarking images, and original watermark image (W) is divided into several arrays of 1 * 8 [w (1), w (2), w (3), w (4), w (5), w (6), w (7), w (8)] according to determined length;

(3b) the 7th and the 8th each array of 1 * 8 is changed to 0, only keeps the information of first six digits, the dope vector that obtains keeping [w (1), w (2), w (3), w (4), w (5), w (6), 0,0];

(3c) each one-dimension array that original watermark image (W) branches away is handled, obtained the dope vector of all reservations, with the dope vector splicing of these reservations, reconstruct the watermarking images (T) that remains again according to step (3b);

(3d) utilize the arnold disorder method that the watermarking images (T) that remains is carried out scramble and handle, obtain watermark (W1) behind the scramble.

4. digital watermarking extracting method based on gray theory comprises following process:

1) image (P) that will contain watermark is divided into four: (i (1,1), i (1,2), i (2,1) and i (2,2)), find the image block (J) that contains watermark;

2) will contain watermarking images piece (J) and carry out wavelet transform, and obtain this and contain watermarking images piece high-frequency sub-band HH, medium-high frequency subband HL, LH, low frequency sub-band LL, and this is contained the low frequency sub-band LL SS symbolic representation of watermarking images piece;

3) utilize the watermark extracting formula $W2=\frac{F2-a2\&times;\mathrm{SS}}{1-a2},$ Extract the watermarking images behind the scramble, and, obtain the watermarking images of disorderly handling through being inverted (Q) the unrest that is inverted of the watermarking images (W2) of this extraction;

4) watermarking images of disorderly handling through being inverted (Q) is carried out gray prediction and handle, recover watermarking images (WW).

5. digital watermarking extracting method according to claim 4, wherein step 4) is carried out according to the following procedure:

4a) being divided into several one-dimension array of 1 * 8 through the watermarking images of disorderly handling (Q) that is inverted, and with gray model G M (1,1) intension type to one of them array [w (1), w (2), w (3), w (4), w (5), w (6), 0,0] carrying out gray prediction handles, dope and be changed to 0 the 7th information w (7) and the 8th information w (8), the one-dimension array that is restored [w (1), w (2), w (3), w (4), w (5), w (6), w (7), w (8)];

4b) according to step 4a) each 1 * 8 the one-dimension array that is divided into through the random watermarking images of handling (Q) that is inverted is handled, and all one-dimension array that recover are spliced, reconstruct the watermarking images (WW) of recovery.

Images