CN1710610A - A digital watermark anti-counterfeiting method against digital-to-analog conversion process - Google Patents

Abstract

The present invention relates to a digital watermark anti-counterfeiting method against digital-analog-to-analog conversion process, which comprises the following steps: a) the watermark embedding process uses a binary digital watermark algorithm to embed a special information-containing watermark in a specific area of a selected protected object Binary digital watermark; b) the watermark extraction decision process is characterized in that: the selected specific area includes a foreground image and a background image, the edge of the selected specific area foreground image is haloed, the foreground image is divided into blocks, and Before embedding the watermark, set the synchronization signal reflecting the geometric characteristics of the foreground image on the foreground image, define two different image block modes to represent the watermark information, carry out the watermark embedding, and use the pixel value variance of the image block to analyze the distribution of the image block pixels when the watermark is extracted and judged. Description, extract the watermark according to a determined threshold and compare the variance of the pixel value of the image block; use the correlation between the extracted watermark sequence and the original watermark sequence to judge whether the watermark information is credible. The invention is beneficial to accurately segment the foreground image and the background area during the preprocessing of extracting the watermark, prevent the original image from being distorted to the greatest extent, better resist printing and scanning attacks, and realize the anti-counterfeiting protection function of the simulation domain.

Description

A digital watermark anti-counterfeiting method against digital-to-analog conversion process

Technical field

The invention relates to a watermark anti-counterfeiting method, in particular to a digital watermark anti-counterfeiting method resistant to digital-analog-analog conversion process.

Background technique

As a novel information hiding technology, digital watermarking provides a new solution to solve a series of problems such as copyright protection, source authentication, tampering authentication, online distribution, user tracking and identity authentication on the open network. But now people mostly focus on the research of digital watermarking technology for digital images, audio and video, and there are few researches on digital watermarking algorithms that use documents and binary images as carriers. The difficulty in the design of the digital watermarking algorithm of the carrier is that the document or binary image can represent the entire image with only two different gray values. There are almost no textures and details, and some are just different geometric shapes. Using such It is difficult for the carrier to design an algorithm that meets the requirements of digital watermarking. However, documents and binary images are widely used, including digital signatures, handwritten signatures, text, etc. The research on its protection issues is of great significance from practice to theory, and now it can provide both the verification function of the digital domain and the There are few examples of anti-counterfeiting functions using documents and binary images as carriers after digital-to-analog and analog-to-digital conversion processing. General electronic encryption technology or digital watermarking algorithms only provide anti-counterfeiting functions in the digital domain for documents, and cannot resist geometric attacks and similar printing. Scanned modulo-digital-analog attacks cannot play an anti-counterfeiting function in the analog domain.

Contents of Invention

The object of the present invention is to provide a digital watermark anti-counterfeiting method that is resistant to digital-analog-analog conversion process. By adopting a binary digital watermark algorithm, a binary value containing special information is embedded in a specific area of the selected protected object. The method of digital watermark makes this specific area have double anti-counterfeiting functions in digital domain and analog domain, which can not only provide verification function in digital domain, but also provide anti-counterfeiting function after digital-analog modulus processing; it can not only protect digital domain documents, but also It can protect paper documents after digital-to-analog conversion; it can realize both online verification and offline anti-counterfeiting; the digital signature generated by this method can approach the traditional signature function, without

It can only be stamped on the blank space of contracts and other documents, and it can also be stamped on complex backgrounds with text and patterns.

A digital watermark anti-counterfeiting method against digital-to-analog conversion process, comprising the following steps:

a) The watermark embedding process uses a binary digital watermark algorithm to embed a binary digital watermark containing special information in a specific area of the selected protected object;

b) Watermark extraction and judgment process

It is characterized by:

Selected specific areas include foreground and background images,

Before the watermark is embedded, the foreground image is haloed, and the halo that is different from the background is added to distinguish the foreground image from the background area, and the foreground image is divided into blocks. When all the pixels of the block are inside the foreground image Embed the watermark information. After block division, a synchronization signal reflecting the geometric characteristics of the foreground image is set in the foreground image to ensure that synchronization can be achieved when extracting the watermark, and to overcome the possible desynchronization introduced in the digital-to-analog/analog-to-digital conversion process similar to printing and scanning. The foreground image is embedded with watermark block by block in raster scanning order according to two block modes. The block mode is divided into two types corresponding to 1/0: the first type does not process the image block, and the second type modifies the image according to the predefined value The pixel value of the central part of the block; vice versa; it can also be realized by darkening the central pixel of the first type of image block and brightening the central pixel of the second type of image block. Watermark information is a binary sequence of important information that needs to be protected. The range of the central portion of the image block depends on the size of the foreground block; after modification, the pixel values of the central portion of the image block gradually change outwards in a circular manner until they are close to the pixel values of the foreground image. In watermark extraction and judgment, for documents that need to be verified or authenticated, the foreground image is first extracted from the background and processed synchronously; the selected specific area is divided by the same block division method as before embedding; the variance of the pixel value of the image block is calculated , compare with a given threshold to extract the watermark, for example, when the variance is greater than the given threshold, the watermark information embedded in the block is judged to be 1, and when the variance is less than or equal to the given threshold, the watermark information embedded in the block is judged to be 0, use correlation judgment to judge the validity of all extracted watermark sequences.

The digital watermark anti-counterfeiting method provided by the present invention is a semi-fragile algorithm in the digital domain, and can resist digital-to-analog (printing, printing, etc.) and modulo-to-digital (scanning, digital photography, etc.) processes in the analog domain. A robust algorithm for emerging geometric attacks.

Beneficial effects of the present invention:

1. By adding halo processing, the foreground image is distinguished from the background area, which is beneficial to accurately segment the foreground image and the background area when extracting the watermark preprocessing.

2. By adding block division, it is guaranteed that the block can be used as a watermark carrier to embed watermark information only when all pixels of the block are inside the foreground image, so as to prevent the distortion of the original image to the greatest extent.

3. Through the setting of the synchronous signal, the problem of geometric deformation introduced in the analog-to-analog conversion process similar to printing and scanning is overcome.

4. By embedding a binary watermark containing special information in a specific area of the selected protected object, a large amount of watermark information can be skillfully embedded while preventing distortion.

5. Use the variance of the pixel value of the image block to describe the pixel distribution of the block, and make a watermark judgment based on this, scientifically, effectively and concisely complete the watermark information extraction and judgment, and have higher robustness to the disturbance of the pixel value, and can Better resist the signal attenuation of the digital-to-analog process, and realize the anti-counterfeiting protection function of digital documents in the analog domain.

Description of drawings

Fig. 1 is the flow chart of the watermark embedding of embodiment 1 of the present invention;

Fig. 2 is the document transmission process protected by the digital watermark anti-counterfeiting method against digital-to-analog conversion process in Embodiment 1 of the present invention;

Fig. 3 is a flow chart of watermark extraction and judgment in Embodiment 1 of the present invention;

Fig. 4 is a schematic diagram of an enterprise official seal not embedded with watermark information in Embodiment 1 of the present invention;

Fig. 5 is a schematic diagram of embedding watermark information enterprise official seal in Embodiment 1 of the present invention;

6 is a schematic diagram of an enterprise official seal embedded with watermark information in a text background in Embodiment 1 of the present invention;

7 is a schematic diagram of an image block that can embed watermark information in a foreground image according to Embodiment 1 of the present invention;

Fig. 8 shows two image block modes corresponding to watermark bit 1 or 0 in Embodiment 1 of the present invention;

Fig. 9 is a schematic diagram of an enterprise official seal not embedded with watermark information in Embodiment 2 of the present invention;

Fig. 10 is a schematic diagram of an enterprise official seal embedded with watermark information in Embodiment 2 of the present invention;

FIG. 11 is a flow chart of watermark embedding in Embodiment 2 of the present invention;

Fig. 12 shows two image block modes corresponding to watermark bit 1 or 0 in Embodiment 2 of the present invention;

Fig. 13 is a document transmission process protected by a digital watermark anti-counterfeiting method against digital-to-analog conversion process in Embodiment 2 of the present invention.

Detailed ways

Below in conjunction with the watermark embedding and watermark extraction judgment process of the company's official seal, the present invention will be further elaborated through the embodiments;

Example 1:

The entire embedding and detection process of the digital watermark anti-counterfeiting method against the digital-to-analog conversion process can be subdivided into the following steps for description: a) preprocessing before watermark embedding; b) watermark embedding; c) possible transmission process of watermarked documents ; d) Preprocessing before watermark extraction; e) Watermark extraction judgment. Fig. 1 shows a) preprocessing before watermark embedding and b) watermark embedding process; Fig. 2 shows the document transmission process protected by the digital watermark anti-counterfeiting method against digital-to-analog conversion process; Fig. 3 contains d) pre-processing before watermark extraction Processing and e) watermark extraction judgment; Figure 4 shows the official seal of the enterprise that does not embed the watermark information, it is the official seal of the enterprise customized according to the geometric size required by the country, and is called the original image; its background is white, and the foreground image (that is, the five-pointed star in the middle , "Tianyin Signature Test Special Seal", words, ring) are red, and r is the pixel value of the red color. In practical applications, the foreground image can also be other single color and shape seals and signature patterns. Fig. 5 shows the official seal of the enterprise after embedding the watermark information. Figure 6 shows the situation where the watermarked stamp is stamped on a complex background. Set the resolution to P pixels per inch, the size of an ordinary corporate seal is D×D, and the radius width of the circumscribed circle of the five-pointed star is not less than l pixels. In this example, set D to 496 and l to 93.

The original watermark information 5 is the binary sequence of important information to be protected W={w _j , j=1, 2,..., M}, w _j ∈ {0, 1}, M is the sequence length, this important The information can include document authenticity information, certificate information, and error correction codes, depending on different applications. In order to improve security, the watermark information is scrambled by the private key and then embedded. The private key has various forms and requires the user or authorization Unique to Fang.

a) Preprocessing before watermark embedding:

As shown in Figure 1, the foreground image 1 of the selected area is the above-mentioned corporate seal in this example, adding a light-colored, imperceptible halo different from the background and foreground colors to the edge of the foreground image, called adding Halo processing 2 is intended to distinguish the foreground image from the background area, which is beneficial to accurately segment the foreground image and the background area when extracting the watermark preprocessing. In this way, when the foreground image is superimposed on the document background, the two are disconnected after binarization, and we can use an algorithm such as seeking region growth to extract the foreground image.

First, add a halo 2 with a width of width pixels on the outside of the five-pointed star and circle on the corporate seal. These pixel values are close to the pixel value s of the background color r.

Then perform block division processing 3, the present invention only divides the foreground image into blocks, intending to embed the watermark inside the foreground image without destroying the edge of the foreground image; the watermark can only be embedded when all pixels of the block are inside the foreground image. As shown in Figure 7, the colored ring is divided into blocks, for the block 19 obtained by the division, all pixels are inside the foreground image, and this block is marked as an area that can be embedded with a watermark; for the block 20 obtained by the division, it Part of the pixels do not belong to the foreground image of the ring, and cannot be used as a block that can be embedded with a watermark, otherwise the edge of the ring may be damaged. The number of block divisions is determined according to the length of the binary watermark sequence and the length of the synchronization signal; the length of the synchronization signal is described below. In this example, the foreground image of the selected area is divided into blocks 3, and the block size is set to d×d. When the pixels of a certain block are all inside the foreground image, this block is marked as an area that can be embedded with a watermark.

After the block is completed, select n (n>=3) image blocks that can represent the geometric characteristics of the foreground image to set the synchronization signal 4 to ensure that the synchronization can be achieved when extracting the watermark, and overcome the synchronization attack that may be introduced during the printing and scanning process, that is, Geometric deformation; while setting the synchronization signal, the watermark capacity of n bits is also sacrificed. Therefore, when the above block is divided into 3, it is required that the embeddable watermark block is at least M+n blocks, and M is the sequence length of the original watermark information 5 . In this example, the blocks in the middle of the upper, lower, left, and right sides are set as sync signal 4 by increasing the brightness of the central area of the block.

In the digital domain, the background area may be a completely blank document gap, or it may be a variety of text with clear edges; in the analog domain, after the signal attenuation of the two complicated processes of printing and scanning, the background area may be a blank corresponding to the digital domain The area has uniform or uneven noise, and it may also be a text background with noise, as shown in Figure 6. The edge halo of the foreground image added in the image preprocessing of the present invention makes it simple and easy for us to separate the foreground image from the background before extracting the watermark. The setting of the synchronous signal is one of the prerequisites for correctly extracting the watermark information.

b) Watermark embedding process:

The watermark embedding process 7 is an important part of the watermark embedding flowchart shown in FIG. 1 . The present invention embeds the original watermark information 5 block by block in a raster scanning manner. As mentioned above, only when all the pixels of the block are inside the foreground image, the block can be used as the carrier of the watermark to embed watermark information.

The embedding of the watermark is reflected by setting different block patterns for different watermark information. This block pattern should be easy to distinguish in the watermark extraction as shown in Figure 3. Fig. 8 shows an example of division of block patterns. There are two types of block modes. The first type keeps the image block as it is without modification21, and the second type uses predefined pixel values to replace the pixel values in the center of the image block22. For example, when w _j is 0, the image block is not changed, and when w _j is 1, the pixel value of the pixel in the center of the image block is set to a value g close to r but different from r, and vice versa.

The extent of the central part of the image block depends on the size of the foreground image block. When it is necessary to modify the pixel value of the central part of the image block, generally the pixel value of the entire central part is g; considering the visual acceptability, the pixel value of the central part is modified in a circular manner so that g gradually approaches the foreground image color r , the quality of the foreground image obtained in this way is very good and easy to accept.

In the embodiment, the region marked as embedding watermark information is scanned in a raster scanning manner, and the original watermark information 5 is scrambled according to the private key 6 and then embedded therein. In this embodiment, two different block modes are set: when the watermark information is bit 1, the corresponding block center pixel value gradually decreases from the inside to the outside and approaches the foreground image pixel value; when the watermark information is bit 0, the corresponding block does not change.

c) The transmission process that the watermarked document may undergo

As mentioned above, Fig. 4 shows the selected area of the document, that is, the corporate official seal in this example. After embedding the watermark, it becomes the image shown in Fig. 5, which has a high visual acceptability. As shown in Figure 2, after the watermark is embedded in the selected area of the document, a document 8 with anti-counterfeiting function is obtained, and the entire document may be printed 9 and delivered in a simulated form of a paper document; when the authenticity of the document needs to be verified, the paper document The document is reconverted through the scanning process 10 into a document 12 to be authenticated in digital form. The document 8 with anti-counterfeiting function may also be directly stored and delivered 11 in the digital domain, and becomes the document 12 to be authenticated.

d) Preprocessing before watermark extraction:

The official seal of the enterprise containing the watermark information, that is, the anti-counterfeit document 8, is printed 9, expressed in the form of a paper document, and scanned 10 to become a document to be authenticated in the digital domain 12, or the official seal of the enterprise containing the watermark information is still in the digital transmission process 11. Documents to be authenticated 12 in the digital domain. The document 12 to be authenticated obtained through the transmission in the analog domain or the digital domain enters the preprocessing before watermark extraction and decision. As shown in FIG. 3 , this step mainly includes foreground image extraction 13 and synchronization recovery 14 . Firstly, use the region growing algorithm and added halo to extract the corporate official seal foreground image 13, and then use the snake crawler algorithm to find the 4 holes with the largest area, and calculate their centroids. It can be considered that the obtained four centroid coordinates are Step a) synchronous signal coordinates set in the preprocessing before watermark embedding; thus perform synchronous recovery 14; perform the same block division 15 on the foreground image as in step a) preprocessing before watermark embedding according to the detected synchronous signal, and the logo may be embedded The block of watermark information, the block inside the foreground image enters the watermark extraction decision step.

e) Watermark Extraction Judgment

As shown in FIG. 3 , by block division 15 , mark the blocks that may embed watermark information, perform watermark information extraction 16 , and descramble the extracted watermark through the private key 17 (ie, 6 in FIG. 1 ).

After the printing/scanning process of the foreground image, even the pixel values of the image block area without pixel modification are no longer consistent with the original pixel values, but various noises are introduced. In order to overcome noise such as pixel value disturbance brought by the two digital-analog/modulus processes of printing/scanning, the present invention uses the image block standard deviation σ to describe the distribution of image block pixels, and according to a given or adaptively determined The threshold T decides whether the watermark embedded in each block is 0 or 1.

The specific process of watermark information extraction 16 is as follows: c) Calculate the standard deviation s _j of the pixel values marked as embedded watermark information blocks in the preprocessing before watermark extraction. When σ _j ≥ T, it is judged that the watermark information w _j embedded in this block is 1 , when σ _j <T, it is judged that the watermark information w _j embedded in this block is 0. When adaptively determining the threshold T, it is necessary to calculate the variance σ _j of all embedded watermark image blocks in the selected area:

${\mathrm{<span\; class="notranslate">\; \&sigma;</span>}}_{\mathrm{<span\; class="notranslate">\; j</span>}}<span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; L</span>}<span\; class="notranslate">\; \&times;</span>\mathrm{<span\; class="notranslate">\; N</span>}}\underset{\mathrm{<span\; class="notranslate">\; m</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}\underset{\mathrm{<span\; class="notranslate">\; no</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}{<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; no</span>}}<span\; class="notranslate">\; -</span>\stackrel{<span\; class="notranslate">\; \&OverBar;</span>}{\mathrm{<span\; class="notranslate">\; x</span>}}<span\; class="notranslate">\; )</span>}^{\mathrm{<span\; class="notranslate">\; 2</span>}}$

The size of the jth image block is L×N, and m/n are row/column numbers of pixels respectively. Let α be a constant determined based on a large number of experimental experiences, and the threshold T adaptively determined as

$\mathrm{<span\; class="notranslate">\; T</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; \&alpha;</span>}\underset{\mathrm{<span\; class="notranslate">\; j</span>}}{\overset{\mathrm{<span\; class="notranslate">\; m</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{\mathrm{<span\; class="notranslate">\; \&sigma;</span>}}_{\mathrm{<span\; class="notranslate">\; j</span>}}<span\; class="notranslate">\; /</span>\mathrm{<span\; class="notranslate">\; m</span>}$

The watermark bits extracted from each block form a watermark sequence information.

The present invention uses the correlation judgment method to judge whether the extracted watermark sequence is valid: suppose the extracted watermark sequence is w', calculate the correlation coefficient ρ=W·W'/M between w' and the original watermark sequence 5W (denotes two vectors), and make a decision based on a large amount of test data and a threshold value ρ _thr deduced from an application-required false alarm probability P _e0 .

${\mathrm{\&rho;}}_{\mathrm{thr}}=\sqrt{2}\mathrm{\&sigma;}\&Center\; Dot;\mathrm{erf}{c}^{-1}\left(2P_{e0}\right),$ where erfc ^-1 is the error complement function $\mathrm{erfc}\left(x\right)=\frac{2}{\sqrt{\mathrm{\&pi;}}}\underset{x}{\overset{\&infin;}{\&Integral;}}{e}^{-t^2}\mathrm{dt}$ The inverse function of , σ is the standard deviation of the foreground image, which is represented by the variance threshold T calculated above in practice:

${\mathrm{<span\; class="notranslate">\; \&rho;</span>}}_{\mathrm{<span\; class="notranslate">\; thr</span>}}<span\; class="notranslate">\; =</span>\sqrt{\mathrm{<span\; class="notranslate">\; 2</span>}}\mathrm{<span\; class="notranslate">\; T</span>}<span\; class="notranslate">\; \&CenterDot;</span>\mathrm{<span\; class="notranslate">\; erf</span>}{\mathrm{<span\; class="notranslate">\; c</span>}}^{<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 2</span>}{\mathrm{<span\; class="notranslate">\; P</span>}}_{\mathrm{<span\; class="notranslate">\; e</span>}\mathrm{<span\; class="notranslate">\; 0</span>}}<span\; class="notranslate">\; )</span>$

Since the variance threshold T calculated from the foreground images obtained through different transmission channels is different, the value of ρ _thr is not constant and is adaptive to the image under test. When ρ≥ρ _thr , w' is consistent with W, and it is judged that the watermark W exists; when ρ<ρ _thr , w' is not consistent with W, it is judged that the watermark W does not exist, and finally a credible judgment result is obtained18.

Example 2:

The entire embedding and detection process of the digital watermark anti-counterfeiting method against the digital-to-analog conversion process can be subdivided into the following steps for description: f) preprocessing before watermark embedding; g) watermark embedding; h) possible transmission process of watermarked documents ; i) Preprocessing before watermark extraction; j) Watermark extraction judgment. Fig. 11 shows f) preprocessing before watermark embedding and g) watermark embedding process; Fig. 13 shows the possible transmission process that the document protected by the digital watermark anti-counterfeiting method of anti-DAC process; i) preprocessing before watermark extraction The processing and j) watermark extraction and judgment process are the same as those in Embodiment 1, as shown in FIG. 3 . Figure 9 shows the corporate official seal without watermark information, which is a corporate official seal customized according to the geometric size required by the country, called the original image; its background is white, and the foreground image (that is, the five-pointed star in the middle, "Tianyin signature test special Chapter " word, circle) is different from embodiment 1 and is blue, and r is the pixel value of this blue. In practical applications, the foreground image can also be other single color and shape stamps and signature patterns. Fig. 10 shows the corporate official seal after embedding watermark information. Set the resolution to P pixels per inch, the size of an ordinary corporate seal is D×D, and the radius width of the circumscribed circle of the five-pointed star is not less than l pixels. In this example, set D to 496 and l to 93.

The original watermark information 27 is a binary sequence W ₌ {w _j , j=1, 2, . This important information can include document authenticity information, document signer identity information and error correction codes, depending on different applications. In order to improve security, the watermark information is scrambled by using the private key 28 and then embedded. The private key comes in various forms, and requires the user or the authorizing party to be unique.

f) Preprocessing before watermark embedding:

As shown in FIG. 11 , the foreground image 23 of the selected area is the above-mentioned official seal of the company in this example.

As shown in FIG. 11, block division processing 24 is performed first. The invention only divides the foreground image into blocks, and intends to embed the watermark inside the foreground image without destroying the edge of the foreground image; only when all the pixels of the block are inside the foreground image can the watermark be embedded. Same as in Embodiment 1, the colored ring is divided into blocks, as shown in Figure 7, for the block 19 obtained by the division, all pixels are inside the foreground image, and this block is marked as an area that can be embedded with a watermark; As far as the obtained block 20 is concerned, some of its pixels do not belong to the foreground image of the ring, and cannot be used as a block that can be embedded with a watermark, otherwise the edge of the ring may be destroyed. The number of block divisions is determined according to the length of the binary watermark sequence and the length of the synchronization signal; the length of the synchronization signal is described below. In this example, the foreground image of the selected area is divided into blocks 24, and the block size is set to d×d. When the pixels of a certain block are all inside the foreground image, this block is marked as a watermark-embeddable area.

After the block is completed, select n (n>=3) image blocks that can represent the geometric features of the foreground image to set the synchronization signal 25, so as to ensure that synchronization can be achieved when extracting the watermark, and overcome the synchronization attack that may be introduced during the printing and scanning process, namely Geometric deformation; while setting the synchronization signal, the watermark capacity of n bits is also sacrificed. Therefore, when the above-mentioned block division 24 is required, the embeddable watermark block is required to be at least M+n blocks, and M is the sequence length of the original watermark information 27 . In this example, the blocks in the middle of the upper, lower, left, and right sides are set as the synchronous signal 25 by increasing the brightness of the central area of the block.

Then, add a light-colored, imperceptible halo on the edge of the foreground image, which is different from the background and foreground colors, which is called halo processing 26. Accurately segment foreground image and background regions during preprocessing. In this way, when the foreground image is superimposed on the document background, the two are disconnected after binarization, and we can use an algorithm such as seeking region growth to extract the foreground image. First, a halo 26 with a width of width pixels is added to the outside of the five-pointed star and the circle of the corporate seal, and these pixel values are close to the pixel value s of the background color r.

In the digital domain, the background area may be a completely blank document gap, or various texts with clear edges; in the analog domain, after the signal attenuation of the two complex processes of printing/printing and digital shooting, the background area may be corresponding to the digital The blank area of the domain has uniform or non-uniform noise, and may also be a text background with noise, as shown in FIG. 6 in Embodiment 1. The edge halo of the foreground image added in the image preprocessing of the present invention makes it simple and easy for us to separate the foreground image from the background before extracting the watermark. The setting of the synchronous signal is one of the prerequisites for correctly extracting the watermark information.

g) Watermark embedding process 29:

The watermark embedding process 29 is an important part of the watermark embedding flowchart shown in FIG. 11 . The present invention embeds the original watermark information 27 block by block according to the raster scanning method. As mentioned above, only when all the pixels of the block are inside the foreground image, the block can be used as the carrier of the watermark to embed watermark information.

The embedding of the watermark is reflected by setting different block patterns for different watermark information. This block pattern should be easy to distinguish in the watermark extraction as shown in Figure 3. Fig. 12 shows an example of division of block patterns. The block modes are divided into two types, the first type darkens 31 pixels in the central part of the image block, and the second type brightens 32 the central part of the image block with predefined pixel values. For example, when w _j is 0, the central part of the image block is darkened; when w _j is 1, the pixel value of the pixel in the central part of the image block is set to a value g close to r but different from r, and vice versa. The difference from Embodiment 1 is that the pixels in the central part of the first type of image block are darkened 31, because the noise introduced in the process of printing/printing and digital photography and scanning is not obvious in the darker part, which can be further overcome Influence of the analog-to-analog conversion process in the example.

The extent of the central part of the image block depends on the size of the foreground image block. When it is necessary to modify the pixel value of the central part of the image block, generally the pixel value of the entire central part is g; considering the visual acceptability, the pixel value of the central part is modified in a circular manner so that g gradually approaches the foreground image color r , the quality of the foreground image obtained in this way is very good and easy to accept.

In the embodiment, the region marked as embedding watermark information is scanned in a raster scanning manner, and the original watermark information 23 is scrambled according to the private key 28 and then embedded therein. In this embodiment, two different block modes are set: when the watermark information is bit 1, the corresponding block center pixel value gradually decreases from the inside to the outside to approach the foreground image pixel value; when the watermark information is bit 0, the corresponding block center pixel value is Gradually increases towards the outside to the foreground image pixel value.

h) The transmission process that the watermarked document may undergo

As mentioned above, Fig. 9 shows the selected area of the document, that is, the corporate official seal in this example. After embedding the watermark, it becomes the image shown in Fig. 10 , which has a high visual acceptability. As shown in Figure 13, after the watermark is embedded in the selected area of the document, it becomes a document 33 with anti-counterfeiting function, and the entire document may be passed through the analog form of the paper document after printing or printing 34; when the authenticity of the document needs to be verified, The paper document is converted back into a digital form of the document to be authenticated 37 through a digital photographing process 35 . For the document 33 with anti-counterfeiting function, the entire document may also be directly stored and passed 36 in the digital domain, and become the document 37 to be authenticated. Anti-counterfeiting documents 33 and documents to be authenticated 37

i) Preprocessing before watermark extraction:

The official seal of the company containing the watermark information, that is, the anti-counterfeit document 8 (33 in Figure 13), is represented in the form of a paper document through printing/printing 34, and becomes a document 37 to be authenticated in the digital domain through digital photography 35, or contains the watermark information enterprise After the digital storage and transmission process 36, the official seal is still a document to be certified 37 in the digital domain, that is, the document to be certified 12 in FIG. 3 . The document 12 to be authenticated obtained through the transmission in the analog domain or the digital domain enters the preprocessing before watermark extraction and decision. As shown in FIG. 3 , this step mainly includes foreground image extraction 13 and synchronization recovery 14 . Firstly, use the region growing algorithm and added halo to extract the corporate official seal foreground image 13, and then use the snake crawler algorithm to find the 4 holes with the largest area, and calculate their centroids. It can be considered that the obtained four centroid coordinates are Step f) synchronous signal coordinates set in the preprocessing before watermark embedding; thus carry out the processing of restoring synchronization 14; according to the detected synchronous signal, the foreground image is divided into blocks 15 identical to step f) preprocessing before watermark embedding, marking Blocks that may embed watermark information, and blocks inside the foreground image enter the watermark extraction decision step.

j) Watermark extraction judgment

Divide by block 15, mark the block that may embed watermark information, perform watermark information extraction 16, and descramble the extracted watermark through private key 17 (ie 28 in Figure 11). After the foreground image undergoes a digital-to-analog/modulus process, even the pixel values of the image block area that have not undergone pixel modification are no longer consistent with the original pixel values, but various noises are introduced. In order to overcome noise such as pixel value disturbance brought by the two processes of modulus/digital-analog, the present invention uses the standard variance σ of the image block to describe the distribution of the pixels of the image block, and according to a given or adaptively determined threshold T Whether the watermark embedded in each block is 0 or 1 is judged.

The specific process of watermark information extraction 16 is as follows: c) the standard deviation s _j of the pixel values marked as embedded watermark information blocks in the preprocessing before watermark extraction, when σ _j ≥ T, the watermark information w _j embedded in the block is judged to be 1, When σ _j <T, it is judged that the watermark information w _j embedded in this block is 0. When adaptively determining the threshold T, it is necessary to calculate the variance σ _j of all embedded watermark image blocks in the selected area:

${\mathrm{<span\; class="notranslate">\; \&sigma;</span>}}_{\mathrm{<span\; class="notranslate">\; j</span>}}<span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; L</span>}<span\; class="notranslate">\; \&times;</span>\mathrm{<span\; class="notranslate">\; N</span>}}\underset{\mathrm{<span\; class="notranslate">\; m</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}\underset{\mathrm{<span\; class="notranslate">\; no</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}{<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; no</span>}}<span\; class="notranslate">\; -</span>\stackrel{<span\; class="notranslate">\; \&OverBar;</span>}{\mathrm{<span\; class="notranslate">\; x</span>}}<span\; class="notranslate">\; )</span>}^{\mathrm{<span\; class="notranslate">\; 2</span>}}$

The size of the jth image block is L×N, and m/n are row/column numbers of pixels respectively. Let α be a constant determined based on a large number of experimental experiences, and the threshold T adaptively determined as

$\mathrm{<span\; class="notranslate">\; T</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; \&alpha;</span>}\underset{\mathrm{<span\; class="notranslate">\; j</span>}}{\overset{\mathrm{<span\; class="notranslate">\; m</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{\mathrm{<span\; class="notranslate">\; \&sigma;</span>}}_{\mathrm{<span\; class="notranslate">\; j</span>}}<span\; class="notranslate">\; /</span>\mathrm{<span\; class="notranslate">\; m</span>}$

The watermark bits extracted from each block form a watermark sequence information.

The present invention uses the correlation judgment method to judge whether the extracted watermark sequence is valid: suppose the extracted watermark sequence is w', and calculate the correlation coefficient ρ=W·W'/M between w' and the original watermark sequence W (denotes two vectors), and make a decision based on a large amount of test data and a threshold value ρ _thr deduced from an application-required false alarm probability P _e0 .

${\mathrm{\&rho;}}_{\mathrm{thr}}=\sqrt{2}\mathrm{\&sigma;}\&Center\; Dot;\mathrm{erf}{c}^{-1}\left(2P_{e0}\right),$ where erfc ^-1 is the error complement function $\mathrm{erfc}\left(x\right)=\frac{2}{\sqrt{\mathrm{\&pi;}}}\underset{x}{\overset{\&infin;}{\&Integral;}}{e}^{-t^2}\mathrm{dt}$ The inverse function of , σ is the standard deviation of the foreground image, which is represented by the variance threshold T calculated above in practice:

${\mathrm{<span\; class="notranslate">\; \&rho;</span>}}_{\mathrm{<span\; class="notranslate">\; thr</span>}}<span\; class="notranslate">\; =</span>\sqrt{\mathrm{<span\; class="notranslate">\; 2</span>}}\mathrm{<span\; class="notranslate">\; T</span>}<span\; class="notranslate">\; \&Center\; Dot;</span>\mathrm{<span\; class="notranslate">\; erf</span>}{\mathrm{<span\; class="notranslate">\; c</span>}}^{<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 2</span>}{\mathrm{<span\; class="notranslate">\; P</span>}}_{\mathrm{<span\; class="notranslate">\; e</span>}\mathrm{<span\; class="notranslate">\; 0</span>}}<span\; class="notranslate">\; )</span>$

Since the variance threshold T calculated from the foreground images obtained through different transmission channels is different, the value of ρ _thr is not constant and is adaptive to the image under test. When ρ≥ρ _thr , w' is consistent with W, and it is judged that the watermark W exists; when ρ<ρ _thr , w' is not consistent with W, and it is judged that the watermark W does not exist. Finally, a convincing verdict was obtained18.

Claims (9)

1. the digital watermark anti-fake method of an anti-digital-to-analog conversion process may further comprise the steps:

A) watermark embed process adopts the bi-level digital watermarking algorithm, embeds a kind of bi-level digital watermark that comprises specific information in selected object of protection specific region;

B) watermark extracting judging process

It is characterized in that:

Selected specific region comprises foreground image and background image,

Selected foreground image edge, specific region is added dizzy the processing,

Foreground image is carried out piece divides, and before watermark embeds, foreground image is provided with the synchronizing signal of reflection foreground image geometric properties, define two kinds of different image block patterns and represent watermark information, carry out watermark and embed,

When adjudicating, watermark extracting is described, according to the extraction watermark of comparing of a definite threshold value and image block pixel value variance with of the distribution of image block pixel value variance to the image block pixel; Whether utilization extracts the correlativity judgement watermark information of watermark sequence and original watermark sequence credible.

2. the digital watermark anti-fake method of anti-digital-to-analog conversion process according to claim 1 is characterized in that, earlier foreground image is added dizzy the processing before watermark embeds, and adds and other halation of background phase region, and it is distinguished mutually with the background area.

3. digital watermark anti-fake method according to claim 1 is characterized in that, foreground image is carried out piecemeal divide, and all pixels of piece are ability embed watermark information when foreground image is inner all.

4. the digital watermark anti-fake method of anti-digital-to-analog conversion process according to claim 1 is characterized in that, after piecemeal is divided, the synchronizing signal of reflection foreground image geometric properties is set at foreground image.

5. the digital watermark anti-fake method of anti-digital-to-analog conversion process according to claim 3, it is characterized in that, to foreground image with the raster scanning sequence block-by-block according to two kinds of block mode embed watermarks, corresponding two classes that are divided into 1/0 of block mode: the first kind does not deal with image block, and second class is revised the pixel value of image block core according to predefined value; Vice versa; Also can pass through the deepening of first kind image block center pixel, the second class image block center pixel blast realizes.

6. the digital watermark anti-fake method of anti-digital-to-analog conversion process according to claim 5 is characterized in that, watermark information is the binaryzation sequence of the important information that needs protection.

7. the digital watermark anti-fake method of anti-digital-to-analog conversion process according to claim 5 is characterized in that, the scope of described image block core depends on the size of prospect piecemeal; The pixel value of revising back image block core outwards gradually changes ringwise to the pixel value near foreground image.

8. the digital watermark anti-fake method of anti-digital-to-analog conversion process according to claim 1 is characterized in that, the document to the needs checking or the evaluation true and false during watermark extracting judgement at first extracts foreground image from background, carry out synchronous processing; With the specific region of selecting being divided with the preceding identical piece division methods of embedding; Computed image piece variance, when variance during greater than this given threshold value, adjudicating the watermark information that this piece embeds is 1, when variance during smaller or equal to this given threshold value, adjudicating the watermark information that this piece embeds is 0.

9. the digital watermark anti-fake method of anti-digital-to-analog conversion process according to claim 8 is characterized in that, uses correlativity to adjudicate the validity of the watermark sequence of judging that all extract.

Images