CN105427231A - SVD dual-layer digital watermarking method capable of avoiding false alarm - Google Patents

Abstract

The invention relates to an SVD (Singular Value Decomposition) dual-layer digital watermarking method capable of avoiding a false alarm. The method comprises: performing SVD on a host image and superposing a singular value matrix of the host image into a watermarking image; performing SVD on the watermarking image superposed with singular values and replacing the singular value matrix of the host image with a singular value matrix of the watermarking image to form an image with a first layer of watermark; performing Hash encryption on an eigenvalue matrix of the watermarking image superposed with the singular values; and performing block discrete cosine transform on the image embedded with the first layer of watermark and embedding a Hash code obtained by encryption into a discrete cosine transform coefficient of each sub-block to obtain an image with a second layer of watermark. The method is suitable for copyright protection of printed products; the dual-layer watermarks can keep high visual quality while increasing watermarking capacity; and the problem of the false alarm happened during watermark extraction is avoided.

Description

A kind of SVD double-layer digital water mark method avoiding police by mistake

Technical field

The present invention relates to image procossing and information security field.The present invention designs a kind of digital watermark embedding and extracting method of anti-print and scan, and the method has raising watermark embedment capacity, reduces the characteristic identifying alert rate by mistake, may be used for protecting literary property of presswork.

Technical background

Universal and fast-developing along with network technology, digital content has no to copy temperately and propagate, and the Copyright Protection of digital product is received much concern, and in order to solve this problem, digital watermark technology arises at the historic moment.Digital watermark technology is a kind of technology embedded by identification information in digital content, and it does not affect former digital content quality substantially, is also not easily found out and revises, but can be extracted identification, for the copyright protection of numerical information and false proof.Through development for many years, digital watermark technology achieves certain achievement.In view of digital watermark technology is realizing in copyright protection and false proof process; numerical information need only be changed; without the need to increasing other costs (as the cost that special paper or specialty inks bring), cause the copyright protection of paper media and the extensive concern of anti-counterfeit field.Digital watermarking be used for paper media copyright protection and false proof time; often to experience printing (printing)-scan process; and print (printing)-scanning and be actually digital content and have passed through D/A and A/D twice conversion; simultaneously also with attacks such as to a certain degree noise, fuzzy, geometric transformations; digital content is caused to be subjected to very large loss or distortion; to such an extent as in digital content, watermark information is difficult to retain, thus lose copyright protection and false proof effect.But; the forgery of the paper media such as current bill, the packing of product, certificate, vital document and copy and bring very large loss to the mechanism such as enterprise, finance, urgently expects that this digital watermark technology without application cost can become and replaces traditional false proof a kind of new technology with copyright protection.Therefore, digital watermarking realizes the false proof and copyright protection of paper media, there is huge applications and is worth, be also faced with very large technological challenge.The robust performance particularly improving the anti-printing of digital watermark technology (printing)-scanning is the basic premise of energy practical application, is a problem being worth research and breaking through.

In recent years, occurred that some can resist the digital watermarking algorithm of printing-scanning attack.Consult data of literatures, roughly have the digital watermarking algorithm of two types to achieve certain effect at present in opposing printing-scanning attack.One class embeds on the basis of the Spectrum Conversions such as discrete cosine transform (DiscreetCosineTransformation, DCT), discrete Fourier transformation (DiscreetFourierTransformation, DFT).Namely region (the being mid-frequency region substantially) embed watermark information that in original image DCT/DFT coefficient, vision is insensitive and basicly stable after printing-scanning attack digital watermarking is sought.Such as, the patent " the anti printing and scanning coloured image blind watermark method based on DCT coefficient statistical property " of the people such as Wang Quan, Chinese patent, the patent No. is CN103559676A.Another kind of is wavelet transform (DiscreetWaveletTransformation, DWT) in conjunction with svd (SingularValueDecomposition, SVD), namely DWT is utilized to obtain the strong low-frequency image of attack resistance stability, again SVD decomposition is carried out to low-frequency image, watermark information is embedded in the singular value coefficient of SVD.Such as, the patent " the Robust Zero watermarking algorithm based on svd, harr wavelet transformation and mean value computation " of Ye Tianyu, Chinese patent, the patent No. is: CN102129659A.The large quantity research carried out through numerous scholar finds, print and scan process is at the applied environment of reality, be not only D/A and A/D conversion in sampling and quantize produce attack, the noise produced due to equipment limitation in print and scan environment, the attack such as fuzzy, can aggravate loss and the distortion of information.Therefore, in actual applications, the attack suffered by the watermark signal of embedding is not one or more the different simple combination of attacking in simple emulation experiment, but the result of Various Complex signal synthesis effect.This makes to be difficult to seek equilibrium point between the robust performance and embed watermark information capacity of visual quality of images, anti-print and scan based on the digital watermark method of DCT.And the equipment in print and scan process, the difference of parameters also also exist very large impact to the robustness of embed watermark information.Comparatively speaking, DWT can obtain the more stable feature of image in conjunction with DWT and SVD in the digital watermark method of SVD, thus obtains more strong robustness, and the capacity of embed watermark information is larger.Therefore, DWT obtains good effect in conjunction with the digital watermark method of SVD in anti-print and scan.But in these methods, in order to increase the visual quality of embed watermark information capacity and raising embedded images, reality is that the singular value that watermark information SVD decomposes is hidden in the singular value of original image DWT changing image substantially.When extracting watermark information, first obtain the watermarking images of scanning and original image DWT convert after singular value difference, recycle the singular vector reconstructuring water-mark information of this unusual value difference and original watermark.But in fact singular vector is only the key message forming any image (as watermarking images), and singular value size just affects the power of final image.Thus, if there is piece image, do not have embed watermark information, after print and scan process, the singular value that its DWT converts has changed, and will obtain watermark information.That is there is no embed watermark, but have identified watermark, thereby produce alert problem by mistake.In the print and scan process of reality, all can there is change in various degree in the singular value of DWT conversion, therefore by mistake alert problem inevitably occurs this type of digital watermark method, and this makes digital watermarking lose the value of the practical application of copyright protection usually.

Summary of the invention

The object of the invention is to, for there is the problem missed alert problem and high power capacity and high-quality and be difficult to balance in existing anti-print and scan digital watermarking algorithm, proposing a kind of SVD double-layer digital water mark method avoiding police by mistake.

The technical solution used in the present invention is, a kind of SVD double-layer digital water mark method avoiding police by mistake, comprises watermark embedment and watermark extracting two processes, implement according to following steps:

First, the embedding of watermark, specifically comprises the following steps:

Steps A 1, utilizes SVD to decompose host image H and watermarking images W, is designated as H=U respectively _hs _hv _h ^tand W=U _ws _wv _w ^t;

Steps A 2, embeds host image singular value matrix S by watermarking images W _h, obtain new image array be calculated as follows:

$I_H^m=S_H+\mathrm{\α}W,$

Wherein, α is embedment strength, and α ∈ [0.01,0.1];

Steps A 3, by the singular value matrix S of host image _hreplace with image the singular value matrix of gained is operated through SVD obtain the image H embedding ground floor watermark information _w, be calculated as follows:

$H_W=U_H{S}_H^m{V}_H^T;$

Steps A 4, by the watermarking images superposing singular value feature value vector matrix U _wand V _wone dimension matrix u is lined up by row _wand v _w, utilize SHA-256 Hash cryptographic algorithm to be encrypted this one dimension matrix, obtain corresponding Hash codes R _uand R _v:

R _U＝Harshing _(SHA-256)(u _W)，

R _V＝Harshing _(SHA-256)(v _W)，

And R _uand R _vbe converted to binary coded form, then to binary R _uand R _vcarry out logic xor operation, obtain a new binary signal R, and get R front 64 as second layer watermark information, be designated as R _w;

Steps A 5, carries out piecemeal DCT operation to the image carrying ground floor watermark information, will second time watermark information R _wbe embedded into direct current (DC) the coefficient c of each sub-block _i(i=1,2 ..., 64) in, wherein step comprises:

A () will carry the image H of ground floor watermark _wbe divided into the sub-block of 8 × 8, DCT operation implemented to each sub-block, obtains the DC coefficient c of each sub-block _i, and calculate its mean value

$\stackrel{\‾}{c}=\underset{i=1}{\overset{64}{\Σ}}{c}_i;$

B () calculates each sub-block DC coefficient c _iwith its mean value between difference:

$diff\left(c_i\right)=c_i-\stackrel{\‾}{c};$

C () is according to second layer watermark information R _wthe bit value R of i position _w(i) and diff (c _i) symbol revise the DC coefficient c of corresponding sub-block _i, concrete operations are: if R _w(i)=1 and diff (c _iduring)>=0, current sub-block DC coefficient c _iincrease | diff (c _i) |; If R _w(i)=1 and diff (c _i) < 0 time, current sub-block DC coefficient c _ireduce | diff (c _i) |; If R _w(i)=0 and diff (c _iduring)>=0, current sub-block DC coefficient c _ito reduce 2 times | diff (c _i) |; If R _w(i)=0 and diff (c _i) < 0 time, current sub-block DC coefficient c _ito increase by 2 times | diff (c _i) |;

D () carries out DCT inverse transformation to each sub-block revising coefficient, obtain the image embedding two-layer watermark information

Then, the extraction of watermark, specifically comprises the following steps:

Step B1, is divided into the sub-block of 8 × 8 by the image H ' of watermark to be extracted and original host image H, and implements DCT operation to each sub-block, obtains the DC coefficient of their sub-blocks separately respectively with (i=1,2 ..., 64), and according to a pair DC coefficient of the two correspondence position sub-block with generate binary signal R ', wherein contain following steps:

A () calculates DC coefficient respectively with mean value respective with it with difference:

$diff\left({c_i}^W\right)=(c_i^W-{\stackrel{\&OverBar;}{c}}^W),$

$diff\left({c_i}^H\right)=(c_i^H-{\stackrel{\&OverBar;}{c}}^H);$

B () from left to right, from top to bottom checks successively with symbol, obtain the binary data R ' of a group 64: when and or and time, bit value R ' (i)=1 of the i position of binary data R '; When and or and time, bit value R ' (i)=0 of the i position of binary data R ';

Step B2, due to original watermark image W and the image superposing host image singular value there is identical eigen vector, therefore watermarking images W is operated the eigenvectors matrix U of gained through SVD _wand V _wone dimension matrix u is lined up by row _wand v _w, and by SHA-256 Hash cryptographic algorithm this one-dimensional signal is encrypted and obtains Hash codes R _uand R _v, then Hash codes is converted to binary coded form, utilizes logic XOR to binary R _uand R _voperate, obtain binary signal R, first 64 that get R, namely obtain the second layer watermark information R of embedding _w;

The second layer watermark information R that the binary signal R ' that step B3, determining step B1 obtain and step B2 obtain _wmatching, if the value of the two is identical, then carries out next step and extract operation, if the value of the two is different, then terminates the operation of extracting watermark, represent in the image of watermark to be extracted and do not carry watermark information;

Step B4, operates the singular value matrix S of gained through SVD by original watermark image W _wthe image H ' replacing with watermark to be extracted operates the singular value matrix S of gained through SVD _{h '}, obtain new image

$I_{H^{\&prime;}}^m=U_W{S}_{H^{\&prime;}}{V_W}^T,$

According to the singular value matrix S of host image _h, image and embedment strength α, extract ground floor watermarking images W ', specific as follows:

$W^{\&prime;}=(I_{H^{\&prime;}}^m-S_H)/\mathrm{\&alpha;}.$

The invention has the beneficial effects as follows, improve based on the high problem of the alert rate of SVD method embed watermark mistiming, attacks of print_scan can be resisted, improve the capacity of watermark embedment, avoid alert phenomenon by mistake to produce, can be used for the copyright protection of printed matter.

Accompanying drawing explanation

Fig. 1 a is the process flow diagram of watermark embed process of the present invention, and Fig. 1 b is the process flow diagram of watermark extraction process of the present invention;

Fig. 2 a is the host image adopted in the embodiment of the present invention, Fig. 2 b is the watermarking images adopted in the embodiment of the present invention, Fig. 2 c is the image after embedding two-layer watermark in the embodiment of the present invention, Fig. 2 d be in the embodiment of the present invention to Fig. 2 c through the printing of KonicaMinoltaC6000 digital printer and Microtek1000XLPlus scanner scanning gained image, Fig. 2 e be in the embodiment of the present invention to Fig. 2 d extract watermarking images.

Embodiment

Below in conjunction with the drawings and specific embodiments, the present invention is described in detail.

Use the present invention in the picture embed watermark method as shown in Figure 1a, concrete steps are as follows:

1st step, utilizes SVD to decompose host image H and watermarking images W, is designated as H=U respectively _hs _hv _h ^tand W=U _ws _wv _w ^t;

2nd step, embeds host image singular value matrix S by watermarking images W _h, obtain new image array be calculated as follows:

$I_H^m=S_H+\mathrm{\&alpha;}W,$

Wherein, α is embedment strength, and gets α=0.01;

3rd step, by the singular value matrix S of host image _hreplace with image the singular value matrix of gained is operated through SVD obtain the image H embedding ground floor watermark information _w, be calculated as follows:

$H_W=U_H{S}_H^m{V}_H^T;$

4th step, by the watermarking images superposing singular value feature value vector matrix U _wand V _wone dimension matrix u is lined up by row _wand v _w, utilize SHA-256 Hash cryptographic algorithm to be encrypted this one dimension matrix, obtain corresponding Hash codes R _uand R _v:

R _U＝Harshing _(SHA-256)(u _W)，

R _V＝Harshing _(SHA-256)(v _W)，

And R _uand R _vbe converted to binary coded form, then to binary R _uand R _vcarry out logic xor operation, obtain a new binary signal R, and get R front 64 as second layer watermark information, be designated as R _w;

5th step, carries out piecemeal DCT operation to the image carrying ground floor watermark information, will second time watermark information R _wbe embedded into the DC coefficient c of each sub-block _i(i=1,2 ..., 64) in, wherein step comprises:

A () will carry the image H of ground floor watermark _wbe divided into the sub-block of 8 × 8, DCT operation implemented to each sub-block, obtains the DC coefficient c of each sub-block _i, and calculate its mean value

$\stackrel{\&OverBar;}{c}=\underset{i=1}{\overset{64}{\&Sigma;}}{c}_i;$

B () calculates each sub-block DC coefficient c _iwith its mean value between difference:

$diff\left(c_i\right)=c_i-\stackrel{\&OverBar;}{c};$

C () is according to second layer watermark information R _wthe bit value R of i position _w(i) and diff (c _i) symbol revise the DC coefficient c of corresponding sub-block _i, concrete operations are: if R _w(i)=1 and diff (c _iduring)>=0, current sub-block DC coefficient c _iincrease | diff (c _i) |; If R _w(i)=1 and diff (c _i) < 0 time, current sub-block DC coefficient c _ireduce | diff (c _i) |; If R _w(i)=0 and diff (c _iduring)>=0, current sub-block DC coefficient c _ito reduce 2 times | diff (c _i) |; If R _w(i)=0 and diff (c _i) < 0 time, current sub-block DC coefficient c _ito increase by 2 times | diff (c _i) |;

D () carries out DCT inverse transformation to each sub-block revising coefficient, obtain the image of the two-layer watermark information of embedded quantity

Use the present invention to extract the method for watermark as shown in Figure 1 b in the image of watermark to be extracted, concrete steps are as follows:

1st step, is divided into the sub-block of 8 × 8 by the image H ' of watermark to be extracted and original host image H, and implements DCT operation to each sub-block, obtains the DC coefficient of their sub-blocks separately respectively with (i=1,2 ..., 64), and according to a pair DC coefficient of the two correspondence position sub-block with generate binary signal R ', wherein contain following steps:

A () calculates DC coefficient respectively with mean value respective with it with difference:

$diff\left({c_i}^W\right)=(c_i^W-{\stackrel{\&OverBar;}{c}}^W),$

$diff\left({c_i}^H\right)=(c_i^H-{\stackrel{\&OverBar;}{c}}^H);$

B () from left to right, from top to bottom checks successively with symbol, obtain the binary data R ' of a group 64: when and or and time, bit value R ' (i)=1 of the i position of binary data R '; When and or and time, bit value R ' (i)=0 of the i position of binary data R ';

2nd step, due to original watermark image W and the image superposing host image singular value there is identical eigen vector, therefore watermarking images W is operated the eigenvectors matrix U of gained through SVD _wand V _wone dimension matrix u is lined up by row _wand v _w, and by SHA-256 Hash cryptographic algorithm this one-dimensional signal is encrypted and obtains Hash codes R _uand R _v, then Hash codes is converted to binary coded form, utilizes logic XOR to binary R _uand R _voperate, obtain binary signal R, first 64 that get R, namely obtain the second layer watermark information R of embedding _w;

3rd step, the second layer watermark information R that the binary signal R ' that determining step B1 obtains and step B2 obtain _wmatching, if the value of the two is identical, then carries out next step and extract operation, if the value of the two is different, then terminates the operation of extracting watermark, represent in the image of watermark to be extracted and do not carry watermark information;

This step but utilizes step below to extract watermarking images when can avoid not having embed watermark, namely avoids the generation of alert phenomenon by mistake;

4th step, operates the singular value matrix S of gained through SVD by original watermark image W _wthe image H ' replacing with watermark to be extracted operates the singular value matrix S of gained through SVD _{h '}, obtain new image

$I_{H^{\&prime;}}^m=U_W{S}_{H^{\&prime;}}{V_W}^T,$

According to the singular value matrix S of host image _h, image and embedment strength α, extract ground floor watermarking images W ', specific as follows:

$W^{\&prime;}=(I_{H^{\&prime;}}^m-S_H)/\mathrm{\&alpha;}.$

For the image shown in Fig. 2, embodiment, illustrates that the present invention is for protecting literary property of presswork.

Host image shown in Fig. 2 and watermarking images are coloured image, when the present invention is applied to coloured image: before embed watermark, first image coloured image being divided into red, green, blue three passages is needed, then utilize the method for embed watermark of the present invention to embed corresponding three-channel watermarking images respectively to three-channel host image, finally synthesize the triple channel image of embed watermark; Before extracting watermark, first the image of watermark to be extracted is divided into the image of red, green, blue three passages, then corresponding three-channel watermarking images is extracted to the method that triple channel image utilizes the present invention to extract watermark respectively, finally the three-channel water watermark image of extraction is synthesized coloured image.

In practical operation: first, copy the watermarking images (image size is 225 × 225) shown in tiling Fig. 2 b, make the size of its host image (image size is 512 × 512) shown in image size with Fig. 2 a identical; Then, use the method for the present invention's embed watermark as shown in Figure 1a, the watermarking images of subchannel is embedded in the host image of respective channel respectively, synthesis passage obtains the image carrying two-layer watermark information, as shown in Figure 2 c, wherein, strength factor α=0.01 of embed watermark; Then, on KonicaMinoltaC6000 digital printer by the image carrying two-layer watermark information shown in Fig. 2 c with the output resolution ratio the output for printing of 600dpi, scan the image of the output for printing again with the scanning resolution of 300dpi with Microtek1000XLPlus scanner, result as shown in Figure 2 d; Finally, use the method for the present invention's extraction watermark as shown in Figure 1 b from Fig. 2 d of subchannel, extract the watermarking images of respective channel, synthesis three-channel water watermark image, extracts result as shown in Figure 2 e.

In order to the beneficial effect of the inventive method is described, calculate this watermarking images extracted and the texture paging index (SSIM) copying the watermarking images be paved into, result is 0.7288.As can be seen here, the image carrying watermark information passes through the attack of printing and scanning process, still can extract the watermarking images showing copyright.

The foregoing is only the embodiment in the present invention, but protection scope of the present invention is not limited thereto.Any those skilled in the art, in the technical scope disclosed by the present invention, can carry out converting or substituting.The conversion carried out and replacement all should be included within the scope of the present invention.Therefore, protection scope of the present invention should be as the criterion with the protection domain of claims.

Claims (3)

1. avoid a by mistake alert SVD double-layer digital water mark method, it is characterized in that, comprise watermark embedment and watermark extracting two processes, implement according to following steps:

First, the embedding of watermark, specifically comprises the following steps:

Steps A 1, utilizes SVD to decompose host image H and watermarking images W, is designated as H=U respectively _hs _hv _h ^tand W=U _ws _wv _w ^t;

Steps A 2, embeds host image singular value matrix S by watermarking images W _h, obtain new image array be calculated as follows:

$I_H^m=S_H+\mathrm{\&alpha;}W,$

Wherein, α is embedment strength, and α ∈ [0.01,0.1];

Steps A 3, by the singular value matrix S of host image _hreplace with image the singular value matrix of gained is operated through SVD obtain the image H embedding ground floor watermark information _w, be calculated as follows:

$H_W=U_H{S}_H^m{V_H}^T;$

Steps A 4, by the watermarking images superposing singular value feature value vector matrix U _wand V _wone dimension matrix u is lined up by row _wand v _w, utilize SHA-256 Hash cryptographic algorithm to be encrypted this one dimension matrix, obtain corresponding Hash codes R _uand R _v:

R _U＝Harshing _(SHA-256)(u _W)，

R _V＝Harshing _(SHA-256)(v _W)，

And R _uand R _vbe converted to binary coded form, then to binary R _uand R _vcarry out logic xor operation, obtain a new binary signal R, and get R front 64 as second layer watermark information, be designated as R _w;

Steps A 5, will carry the image H of ground floor watermark information _wbe divided into the sub-block of 8 × 8, and DCT operation is implemented, by second layer watermark information R to each sub-block _wbe embedded into direct current (DC) the coefficient c of each sub-block _i(i=1,2 ..., 64) in, carry out the image that DCT inverse transformation obtains carrying two-layer watermark information

Then, the extraction of watermark, specifically comprises the following steps:

Step B1, is divided into the sub-block of 8 × 8 by the image H ' of watermark to be extracted and original host image H, and implements DCT operation to each sub-block, obtains the DC coefficient of their sub-blocks separately respectively with and according to a pair DC coefficient of the two correspondence position sub-block with generate binary signal R ';

Step B2, due to original watermark image W and the image superposing host image singular value there is identical eigen vector, therefore watermarking images W is operated the eigenvectors matrix U of gained through SVD _wand V _wone dimension matrix u is lined up by row _wand v _w, and by SHA-256 Hash cryptographic algorithm this one-dimensional signal is encrypted and obtains Hash codes R _uand R _v, then Hash codes is converted to binary coded form, utilizes logic XOR to binary R _uand R _voperate, obtain binary signal R, first 64 that get R, namely obtain the second layer watermark information R of embedding _w;

The second layer watermark information R that the binary signal R ' that step B3, determining step B1 obtain and step B2 obtain _wmatching, if the value of the two is identical, then carries out next step and extract operation, if the value of the two is different, then terminates the operation of extracting watermark, represent in the image of watermark to be extracted and do not carry watermark information;

Step B4, operates the singular value matrix S of gained through SVD by original watermark image W _wthe image H ' replacing with watermark to be extracted operates the singular value matrix S of gained through SVD _h., obtain new image

$I_{H^{\&prime;}}^m=U_W{S}_{H^{\&prime;}}{V_W}^T,$

According to the singular value matrix S of host image _h, image and embedment strength α, extract ground floor watermarking images W ', specific as follows:

$W^{\&prime;}=(I_{H^{\&prime;}}^m-S_H)/\mathrm{\&alpha;}.$

2. a kind of SVD double-layer digital water mark method avoiding police by mistake according to claim 1, is characterized in that: in described steps A 5, by second layer watermark information R _wbe embedded into image each sub-block DC coefficient c _i(i=1,2 ..., 64) operation, concrete grammar is as follows:

First, each sub-block DC coefficient c is calculated _iwith its mean value between difference diff (c _i):

$diff\left(c_i\right)=c_i-\stackrel{\&OverBar;}{c},$

$\stackrel{\&OverBar;}{c}=\underset{i=1}{\overset{64}{\&Sigma;}}{c}_i;$

Then, according to second layer watermark information R _wthe bit value R of i position _w(i) and diff (c _i) symbol revise the DC coefficient c of corresponding sub-block _i, concrete operations are: if R _w(i)=1 and diff (c _iduring)>=0, current sub-block DC coefficient c _iincrease | diff (c _i) |; If R _w(i)=1 and diff (c _i) < 0 time, current sub-block DC coefficient c _ireduce | diff (c _i) |; If R _w(i)=0 and diff (c _iduring)>=0, current sub-block DC coefficient c _ito reduce 2 times | diff (c _i) |; If R _w(i)=0 and diff (c _i) < 0 time, current sub-block DC coefficient c _ito increase by 2 times | diff (c _i) |.

3., according to a kind of SVD double-layer digital water mark method avoiding police by mistake according to claim 1, it is characterized in that: in described step B1, according to DC coefficient with generate the operation of binary signal R ', concrete grammar is as follows:

First, DC coefficient is calculated respectively with mean value respective with it with difference:

$diff\left({c_i}^W\right)=(c_i^W-{\stackrel{\&OverBar;}{c}}^W),$

$diff\left({c_i}^H\right)=(c_i^H-{\stackrel{\&OverBar;}{c}}^H);$

Then, from left to right, check diff (c successively from top to bottom _i ^w) and diff (c _i ^h) symbol, obtain the binary data R ' of a group 64: as diff (c _i ^w)>=0 and diff (c _i ^h)>=0 or diff (c _i ^w) < 0 and diff (c _i ^h) < 0 time, bit value R ' (i)=1 of the i position of binary data R '; As diff (c _i ^w) < 0 and diff (c _i ^h) > 0 or diff (c _i ^w) > 0 and diff (c _i ^h) < 0 time, bit value R ' (i)=0 of the i position of binary data R '.