CN101645162A - Printing-scanning resistance zero-watermarking algorithm - Google Patents

Abstract

A printing-scanning process relates to D/A and A/D double conversation which brings about interferences such as random noise addition, cutting, scaling, rotating and the like; therefore, a printing-scanning attacked resistance digital watermarking algorithm has very high requirement on the robustness. Aiming at the problems, the invention provides a printing-scanning resistance zero-watermarking algorithm by utilizing energy aggregation effect of discrete wavelet transformation low frequency approximation sub-band and singular value stability of singular value decomposition. Partitioning singular value decomposition is carried out on a discrete wavelet transformation low frequency approximation sub-band of an original image; and zero watermarking is generated by judging the size relationship of maximum singular values of two adjacent sub-blocks. The invention can resist printing-scanning attack.

Description

A kind of printing-scanning resistance zero-watermarking algorithm

Technical field

The present invention relates to a kind of field of multimedia signal processing, especially the digital image copyright protection field.The invention provides a kind of printing-scanning resistance zero-watermarking algorithm, the digital picture of needs experience printing-scanning attack is carried out copyright protection.

Background technology

Along with Internet development, increasing digital product wide-scale distribution on network.In many occasions, the digital picture on the network need be printed and rescan into digital picture after treatment again and on network, propagate.How the digital picture of needs experience printing-scanning process is carried out copyright protection and just become a problem of being badly in need of solution.Printing-scanning relates to D/A and twice conversion of A/D, can bring interference such as adding random noise, shearing, convergent-divergent, rotation.The digital watermarking algorithm of anti-printing-scanning requires that these a series of attacks are all had very high robustness.The watermarking algorithm of the anti-printing-scanning attack of research has very high theoretical research and actual application value.

Up to the present, the robustness of most digital watermarking algorithms is not also enough resisted printing-scanning attack, and the digital watermarking algorithm that can resist printing-scanning attack is actually rare.Consult data of literatures, domestic researcher has roughly proposed the digital watermarking algorithm of following anti-printing-scanning attack at present, sees the listed list of references in end, this part.People such as Niu Shaozhang, Wu Hongtao [1] classifies to the discrete cosine transform domain coefficient, by sign quantitaes watermark information in every class.People [2] such as Niu Shaozhang, button heart Xin are by adjusting the order embed watermark information of discrete cosine transform domain intermediate frequency coefficient, utilize the maximum membership grade principle of Fuzzy Pattern Recognition to differentiate watermark information twice during extraction.People such as Sun Yunfeng, Zhai Hongchen [3] utilizes the coloured image digital watermarking system of a kind of anti-printing of Fourier's calculation holographic technical design-scanning, has bigger watermark and embeds capacity.Because the Walsh function has strict orthogonal, people such as Ge Yunlu, Nian Guijun [4] at first utilizes the Walsh function that original watermark is carried out coded modulation, approach the sub-band adaptive embed watermark information in conjunction with human-eye visual characteristic at the wavelet transform low frequency then, realize the embedding and the extraction of the meaningful watermark information of high capacity.People [5] such as Tan is just discussing, Zhu Ningbo propose a kind of anti-printing-scanning digital digital watermark that is applicable to file and picture, by adjusting the wavelet transform territory detail subbands sign quantitaes watermark information of file and picture piece.Seek the forward and backward maintenance as far as possible of digital picture printing-scanning and stablize the crucial part that constant characteristic quantity is this type of algorithm of design.

The wavelet transform low frequency approaches most energy that subband has been assembled original image, has stronger anti-interference stability than other three detail subbands.The singular value that obtains after svd disturbance to external world has good stability.Zero watermarking algorithm often utilizes the invariant feature of initial carrier to produce zero watermark, effectively solves the contradiction between invisibility and the robustness.The key that zero watermarking algorithm has strong robustness is that the characteristic quantity that is utilized has stiff stability to external interference.The present invention utilizes the wavelet transform low frequency to approach the energy accumulating effect of subband and the singular value stability of svd; propose a kind of printing-scanning resistance zero-watermarking algorithm, be intended to realize the digital picture of needs experience printing-scanning attack is carried out copyright protection.

List of references

[1] Niu Shaozhang, Wu Hongtao thanks to trace, Liu Xin, Yang Yixian. the robustness of anti printing and scanning digital watermarking algorithm [J]. and Zhongshan University's journal (natural science edition), 2004,43 (supplementary issue 2): 1-4.

[2] Niu Shaozhang, button heart Xin, Yang Yixian, Hu Wenqing. Data Hiding Algorithm for Halftone Images [J]. electronic letters, vol, 2004,32 (7): 1180-1183.

[3] Sun Yunfeng, Zhai Hongchen, Yang Xiaoping, Wang Mingwei. the application [J] of Fourier calculation holographic digital watermarking in the coloured image antiforge. photoelectron. laser, 2008,19 (7): 952-955.

[4] Ge Yunlu, Nian Guijun, Wang Shuxun. based on the code modulated anti printing and scanning digital watermarking algorithm of Walsh [J]. Southeast China University's journal (natural science edition), 2007,37 (supplementary issue I): 197-200.

[5] Tan is just discussing, Zhu Ningbo, Sun Xingming. based on the anti-printing-scanning digital digital watermark [J] of file and picture. and computer utility research, 2007,24 (12): 199-200,265.

Summary of the invention

The purpose of this invention is to provide a kind of printing-scanning resistance zero-watermarking algorithm, the digital picture of needs experience printing-scanning attack is carried out copyright protection so that realize.

For reaching this purpose, the present invention utilizes the wavelet transform low frequency to approach the energy accumulating effect of subband and the singular value stability of svd, designs a kind of printing-scanning resistance zero-watermarking algorithm.Copyright authentication process flow diagram of the present invention comprises following six kinds of modules:

A, zero watermark generation module;

B, post-processing module;

C, zero watermark extracting module;

D, similarity evaluation module;

E, threshold value are provided with module;

F, copyright ownership judge module.

Modules A comprises following content:

Original image carries out n layer scattering wavelet transformation, and n layer low frequency approximator band is carried out the piecemeal svd, produces zero watermark w by the magnitude relationship of judging two adjacent sub-blocks maximum singular values.

Module B further comprises following content:

The digital picture of printing-scanning back input is carried out operations such as cutting, rotation correction, adjustment image sampling resolution, adjustment image size

Module C comprises following content:

N layer scattering wavelet transformation low frequency approximator band to input picture carries out the piecemeal svd, extracts zero watermark w ' by the magnitude relationship that compares two adjacent sub-blocks maximum singular values.

Module D comprises following content:

Similarity s between the zero watermark sequence w ' of original zero watermark sequence w and extraction is defined as:

$s=1-\left(\underset{j=1}{\overset{K/2}{\mathrm{\Σ}}}{w}_j{w}_j^{\′}\right)/(K/2),$ Wherein

Be XOR.The similarity evaluation module is used for weighing the correlativity between the different sequences.

Module E further comprises following content:

According to initial zero watermark and lock-out attack the similarity between zero watermark of extracting the back, initial zero watermark and at random 0, the similarity setting threshold th between the 1} sequence.

Module F further comprises following content:

Similarity and threshold decision copyright ownership according to forward and backward zero watermark of digital picture printing-scanning.

The present invention realizes that for the digital picture of experience printing-scanning attack copyright protection provides a new way.Wavelet transform low frequency approximator band to original image carries out the piecemeal svd, produces zero watermark by the magnitude relationship of judging two adjacent sub-blocks maximum singular values.Behind digital picture experience printing-scanning attack, similarity that still can be higher is extracted zero watermark, thereby can carry out copyright protection to the digital picture of experience printing-scanning attack.

Description of drawings

Fig. 1 is the copyright authentication process flow diagram;

Among Fig. 2, (a) be original Lena image; (b) be original Peppers image; (c) be original Baboon image;

Among Fig. 3, (a) be the zero watermark of Lena image with at random 0, the similarity between the 1} sequence; (b) be the zero watermark of Peppers image with at random 0, the similarity between the 1} sequence; (c) be the zero watermark of Baboon image with at random 0, the similarity between the 1} sequence;

Among Fig. 4, (a) be the Lena image of experience printing-scanning attack and aftertreatment; (b) be the Peppers image of experience printing-scanning attack and aftertreatment; (c) be the Baboon image of experience printing-scanning attack and aftertreatment.

Embodiment

Below in conjunction with drawings and Examples description is done into-gone on foot to technical scheme of the present invention.

Fig. 1 is the copyright authentication process flow diagram, comprises following process:

(1) zero watermark produces

Original image carries out n layer scattering wavelet transformation, and n layer low frequency approximator band is carried out the piecemeal svd, produces zero watermark w by the magnitude relationship of judging two adjacent sub-blocks maximum singular values.If that is: ${\mathrm{\&sigma;}}_{2\&times;j-1}^1\&GreaterEqual;{\mathrm{\&sigma;}}_{2\&times;\mathrm{<figure-callout\; id="1"\; label="j"\; filenames="A2009100922760013H1.png"\; state="\{\{state\}\}">j</figure-callout>}}^1,$ W then _j=0; Otherwise, w _j=1 $(j=\mathrm{1,2},L,\frac{K}{2}).$ Wherein, σ _t ¹Be the maximum singular value of t sub-piece, w _jBe the j bit watermark of w, K is sub-piece sum.Preserve initial zero watermark, be used for similarity evaluation module one, two, three and calculate similarity between it and other sequences.

(2) aftertreatment

The digital picture of printing-scanning back input is carried out operations such as cutting, rotation correction, adjustment image sampling resolution, adjustment image size, for the zero watermark behind the extraction printing-scanning attack is prepared.

(3) zero watermark extracting

N layer scattering wavelet transformation low frequency approximator band to input picture carries out the piecemeal svd, extracts zero watermark w ' by the magnitude relationship that compares two adjacent sub-blocks maximum singular values.If that is: ${\mathrm{\&sigma;}}_{2\&times;j-1}^{\&prime;1}\&GreaterEqual;{\mathrm{\&sigma;}}_{2\&times;j}^{\&prime;1},$ W ' then _j=0; Otherwise, w ' _j=1 (j=1,2, L, K/2).Zero watermark extracting module has two: module one is used for from suffering that the image that lock-out is attacked extracts watermark sequence; Module two be used for from printed-image of scanning attack extracts watermark sequence.

(4) similarity evaluation

Similarity s between the zero watermark sequence w ' of original zero watermark sequence w and extraction is defined as: $s=1-(\underset{j=1}{\overset{K/2}{\mathrm{\&Sigma;}}}{w}_j\&CirclePlus;w_j^{\&prime;})/(K/2),$ Wherein

Be XOR.The similarity evaluation module has three: module one is used for weighing the correlativity between the zero watermark of extracting after initial zero watermark and the lock-out attack, and the robustness that evaluation algorithms opposing lock-out is attacked provides reference for threshold value is set; Module two is used for weighing initial zero watermark with { 0, the correlativity between the 1} sequence provides reference for threshold value is set at random; Module three is used for weighing the correlativity between the zero watermark of extracting behind initial zero watermark and the printing-scanning attack, and the robustness of evaluation algorithms opposing printing-scanning attack is for judging that copyright ownership provides reference.

(5) threshold value setting

Initial zero watermark that calculates according to similarity evaluation module one and lock-out attack initial zero watermark that similarity, similarity evaluation module two between zero watermark of extracting the back calculate and at random 0, the similarity setting threshold th between the 1} sequence.

(6) copyright ownership is judged

The threshold decision copyright ownership that similarity between the zero watermark of extracting behind initial zero watermark that calculates according to similarity evaluation module three and the printing-scanning attack and threshold value are provided with module settings.When s greater than threshold value th, copyright ownership is in the plaintiff.

Embodiment:

Lena, Peppers, Baboon are that size is 512 * 512 256 grey-level standard images, have different textures, see Fig. 2 (a)-(c) respectively.Image is carried out 3 layer scattering wavelet transformations, is wavelet basis with the harr small echo, and the branch block size that low frequency approaches subband is 2 * 2, and zero watermark sequence length is 512 bits.Weigh vision difference between original image and the attack graph picture with PSNR, with similarity s weigh original zero watermark with at random 0,1} sequence and attack correlativity between the zero watermark of back extraction.

(1) robustness of opposing lock-out attack

1) rotation is attacked

Experiment parameter setting and the results are shown in Table 1.

Table 1 rotation

2) the skew ranks are attacked

Offset column is meant that entire image moves right to the right, and originally several row of rightmost shift out and lose, and several row completions of the Far Left of vacating are black.Experiment parameter setting and the results are shown in Table 2.

Table 2 is offset column to the right

Downward bias is divided a word with a hyphen at the end of a line and is meant that entire image moves down, and several bottom originally row shift out to be lost, and several capable completions in the top of vacating are black.Experiment parameter setting and the results are shown in Table 3.

Table 3 downward bias is divided a word with a hyphen at the end of a line

3) attack at random deletes both in the middle of

Middle delete columns at random is meant from the right first row of deleted row and begins to be moved to the left by row that several row completions of the rightmost of vacating are black.Experiment parameter setting and the results are shown in Table 4.

The middle delete columns at random of table 4

The middle row of deletion at random is meant from bottom's first row beginning of deleted row and moves up line by line that several bottom capable completions of vacating are black.Experiment parameter setting and the results are shown in Table 5.

Delete row in the middle of the table 5 at random

4) size scaling

Experiment parameter setting and the results are shown in Table 6.

Table 6 size scaling

According to table 1-6, for the image of Lena, Peppers, Baboon three width of cloth different texture, algorithm is attacked several lock-outs all has very strong robustness.

(2) threshold value setting

Produce 99 at random and obey equally distributed { 0, the 1} sequence, calculate respectively zero watermark that Lena, Peppers, Baboon three width of cloth images produce and these 0, the similarity between the 1} sequence, the results are shown in Figure 3 (a)-(c), wherein the 50th sequence is the zero watermark that three width of cloth images produce separately.According to Fig. 3 (a)-(c), the zero watermark that three width of cloth images produce and these 0, near mostly fluctuation within a narrow range 0.5 of the similarity between the 1} sequence.Opposing lock-out in conjunction with 3.1 parts is attacked experimental result, and we select 0.75 as threshold value th.

(3) robustness of opposing printing-scanning attack

At first use model digital picture to be printed on common printer paper as the laser printer of Canon L11121E; Use the scanner scanning input of model as CanoScan LiDE 100 then, resolution is set to 400dpi; Then the digital picture of scanning input is carried out aftertreatment, comprise and use Photoshop software that digital picture is carried out cutting and rotation correction, the sampling resolution of image is readjusted into 72dpi, used the bilinear method of interpolation that the image size is readjusted into 512 * 512 etc.Digital picture through printing-scanning, aftertreatment is seen Fig. 4 (a)-(c) respectively.PSNR between they and the original image is respectively 19.3572,18.0845,15.7371, and the similarity between zero watermark is respectively 0.8672,0.8691,0.8594, therefore thinks that the present invention can resist printing-scanning attack.

Claims (7)

1, a kind of printing-scanning resistance zero-watermarking algorithm is used for the copyright authentication field that needs experience the digital picture of printing-scanning attack, comprises following six kinds of modules:

A, zero watermark generation module;

B, post-processing module;

C, zero watermark extracting module;

D, similarity evaluation module;

E, threshold value are provided with module;

F, copyright ownership judge module.

2, a kind of printing-scanning resistance zero-watermarking algorithm according to claim 1, modules A further comprises following content:

Original image carries out n layer scattering wavelet transformation, and n layer low frequency approximator band is carried out the piecemeal svd, produces zero watermark w by the magnitude relationship of judging two adjacent sub-blocks maximum singular values.If that is: ${\mathrm{\&sigma;}}_{2\&times;j-1}^1\&GreaterEqual;{\mathrm{\&sigma;}}_{2\&times;j}^1,$ W then _j=0; Otherwise, $w_j=1(j=\mathrm{1,2},L,\frac{K}{2}).$ Wherein, σ _t ¹Be the maximum singular value of t sub-piece, w _jBe the j bit watermark of w, K is sub-piece sum.Preserve initial zero watermark, be used for similarity evaluation module one, two, three and calculate similarity between it and other sequences.

3, a kind of printing-scanning resistance zero-watermarking algorithm according to claim 1, module B further comprises following content:

The digital picture of printing-scanning back input is carried out operations such as cutting, rotation correction, adjustment image sampling resolution, adjustment image size, for the zero watermark behind the extraction printing-scanning attack is prepared.

4, a kind of printing-scanning resistance zero-watermarking algorithm according to claim 1, module C further comprises following content:

N layer scattering wavelet transformation low frequency approximator band to input picture carries out the piecemeal svd, extracts zero watermark w ' by the magnitude relationship that compares two adjacent sub-blocks maximum singular values.If that is: ${\mathrm{\&sigma;}}_{2\&times;j-1}^{\&prime;1}\&GreaterEqual;{\mathrm{\&sigma;}}_{2\&times;j}^{\&prime;1},$ W ' then _j=0; Otherwise, w ' _j=1 (j=1,2, L, K/2).Zero watermark extracting module has two: module one is used for from suffering that the image that lock-out is attacked extracts watermark sequence; Module two be used for from printed-image of scanning attack extracts watermark sequence.

5, a kind of printing-scanning resistance zero-watermarking algorithm according to claim 1, module D further comprises following content:

Similarity s between the zero watermark sequence w ' of original zero watermark sequence w and extraction is defined as:

$s=1-(\underset{j=1}{\overset{K/2}{\mathrm{\&Sigma;}}}{w}_j\&CirclePlus;w_j^{\&prime;})/(K/2),$ Wherein

Be XOR.The similarity evaluation module has three: module one is used for weighing the correlativity between the zero watermark of extracting after initial zero watermark and the lock-out attack, and the robustness that evaluation algorithms opposing lock-out is attacked provides reference for threshold value is set; Module two is used for weighing initial zero watermark with { 0, the correlativity between the 1} sequence provides reference for threshold value is set at random; Module three is used for weighing the correlativity between the zero watermark of extracting behind initial zero watermark and the printing-scanning attack, and the robustness of evaluation algorithms opposing printing-scanning attack is for judging that copyright ownership provides reference.

6, a kind of printing-scanning resistance zero-watermarking algorithm according to claim 1, module E further comprises following content:

Initial zero watermark that calculates according to similarity evaluation module one and lock-out attack initial zero watermark that similarity, similarity evaluation module two between zero watermark of extracting the back calculate and at random 0, the similarity setting threshold th between the 1} sequence.

7, a kind of printing-scanning resistance zero-watermarking algorithm according to claim 1, module G further comprises following content:

The threshold decision copyright ownership that similarity between the zero watermark of extracting behind initial zero watermark that calculates according to similarity evaluation module three and the printing-scanning attack and threshold value are provided with module settings.When s greater than threshold value th, copyright ownership is in the plaintiff.

Images