CN111861846A - Electronic document digital watermark processing method and system - Google Patents

Abstract

The invention relates to a digital watermark processing method and system for electronic documents. The method includes a watermark embedding step and a watermark extraction step. The watermark embedding step includes: acquiring user rights; collecting watermark information, where the watermark information includes user information; The watermark information is uploaded to the server; the watermark information is received and stored in the database; a plurality of features in the watermark information are extracted according to a predetermined algorithm, and the features are fused to generate a digital watermark; after the electronic document is converted into an image, the The digital watermark is embedded in the image, and the image is restored into an electronic document; the watermark extraction step is a reverse operation of the watermark embedding step. The invention can not only solve the problems of low embedding capacity and poor robustness in the prior art, but also has the advantages of better encryption effect, large key space, strong sensitivity, and resistance to exhaustive attacks and statistical attacks.

Description

Electronic document digital watermark processing method and system

technical field

The invention relates to the technical field of digital watermark processing, in particular to a method and system for digital watermark processing of electronic documents.

Background technique

The rapid development of computer networks and information technology has accelerated the development of office automation and e-commerce. A large number of digital information such as videos, images, emails, documents, etc. are transmitted through the network, but the subsequent illegal infringement and copyright protection problems have become increasingly prominent. Therefore, digital watermarking technology has received extensive attention and research. The realization of digital watermarking technology in multimedia data mainly depends on the existence of a large amount of redundant information in this carrier information, and the data redundancy of electronic document images is very limited compared to other multimedia data, which also makes digital image-based digital images. The development of watermarking technology is relatively limited, and there are problems such as low embedding capacity and poor robustness.

SUMMARY OF THE INVENTION

Based on this, it is necessary to provide a digital watermark processing method and system for electronic documents, which can not only solve the above problems, but also have better encryption effect, large key space, and strong sensitivity, and can resist exhaustive attacks and statistical attacks. Etc.

In order to achieve the above purpose of the invention, the present invention adopts the following technical solutions.

The present invention first provides a digital watermark processing method for electronic documents, including a watermark embedding step, the watermark embedding step specifically includes:

Obtain user permissions;

Collect watermark information, the watermark information includes user information;

uploading the watermark information to the server;

Receive the watermark information and store it in the database;

Extract multiple features in the watermark information according to a predetermined algorithm, and perform feature fusion to generate a digital watermark;

After the electronic document is converted into an image, the digital watermark is embedded in the image, and the image is restored into an electronic document.

In the above-mentioned electronic document digital watermark processing method, it also includes a watermark extraction step, and the watermark extraction step specifically includes:

Convert digitally watermarked electronic documents into images;

Process the image according to a predetermined algorithm to extract the digital watermark;

Using an inverse operation, the digital watermark is converted into a watermarked image.

In the above electronic document digital watermark processing method, the watermark information is the biometric information of the user, including one or more of the user's facial features, hand features, fingerprint features and voice features.

The present invention further provides a digital watermark processing system for electronic documents, comprising a watermark embedding unit, and the watermark embedding unit specifically includes:

Authorization module, used to obtain user permissions;

an information collection module for collecting watermark information, where the watermark information includes user information;

an information uploading module for uploading the watermark information to the server;

a database access module for receiving the watermark information and storing it in the database of the server;

a watermark processing module, used for extracting multiple features in the watermark information according to a predetermined algorithm, and performing feature fusion to generate a digital watermark;

The watermark embedding module is used to embed the digital watermark in the image after converting the electronic document into an image, and then restore the image into an electronic document.

The electronic document digital watermark processing system further includes a watermark extraction unit, and the watermark extraction unit specifically includes:

The conversion module is used to convert the digital watermark-embedded electronic document into an image;

an extraction module, configured to process the image according to a predetermined algorithm to extract the digital watermark;

The restoration module is used for converting the digital watermark into a watermark image by inverse operation.

In the above electronic document digital watermark processing system, the electronic document digital watermark processing system consists of a mobile client and a server, wherein the mobile client includes the authorization module, the information collection module and the information upload module, the server The terminal includes the database, a database access module, a watermark processing module, a watermark embedding module, a watermark decryption module, a search module and an extraction module.

In the above-mentioned electronic document digital watermark processing system, the watermark information is the biometric information of the user, including one or more of the user's facial features, hand features, fingerprint features and voice features;

The mobile client is installed on a mobile phone, and the information collection module includes a camera, a microphone, a fingerprint identifier and/or a face identifier of the mobile phone.

The present invention also provides a digital watermark processing method for an electronic document, including a watermark embedding step of embedding a watermark image in the electronic document, and the watermark embedding step specifically includes:

Convert electronic documents into carrier images with a bit depth of 8;

Select the white pixels to reconstruct the carrier image, that is, use the discrete wavelet algorithm to decompose the carrier image into four frequency domains: LL, HL, LH, HH, and select the HH frequency domain part as the watermark embedding area;

performing singular value decomposition on the watermark image;

Embed the watermark image into the watermark embedding area, that is, replace the singular value of the watermark embedding area with the singular value of the watermark image, and use the inverse operation to obtain the reconstructed carrier sub-image embedded in the watermark;

The reconstructed carrier sub-image is restored to the electronic document image according to the original pixel position.

In the above-mentioned electronic document digital watermark processing method, it also includes a watermark extraction step, and the watermark extraction step specifically includes:

Extracting the pixel points of the watermark-embedded electronic document image to obtain the watermark-embedded carrier image;

Discrete wavelet transform is performed on the carrier image embedded in the watermark, and the HH frequency domain part is selected as the watermark extraction area;

Perform singular value decomposition on the watermark extraction area, and extract the singular values of the watermark extraction area;

Using the inverse operation, the watermark image is restored by using the obtained singular values.

In the above-mentioned digital watermark processing method for electronic documents, the watermark image includes the biometric information of the user, including one or more of the user's facial features, hand features, fingerprint features and voice features.

Compared with the traditional digital watermarking technology, the present invention has the following outstanding advantages:

1. Aiming at the characteristics of pixel distribution of electronic document images, the present invention proposes a digital watermark processing method and system for electronic documents, which utilizes the existence of a large number of white pixel points in electronic documents, which are interspersed between the lines. The characteristics of natural images that do not exist, through these pixels, a new carrier image is reconstructed, the watermark is embedded in it, and then restored to the original pixel position. This method of reconstruction and restoration also realizes the scrambling of the embedded watermark. The watermark embedding method of the present invention operates in the image frequency domain, and combines discrete wavelet transform and singular value decomposition technology to achieve watermark embedding. The experimental results show that the method has good robustness.

2. By analyzing the capacity of white pixels in the electronic document, the present invention can realize 2 to 3 kinds of biological features as watermark images. It is proved by experiments that the embedding of multiple watermarks does not have a great impact on the quality of the carrier image, and the watermarking algorithm is still It has better invisibility and robustness.

Description of drawings

1 is a schematic flowchart of a watermark embedding step in Embodiment 1 of the present invention;

2 is a schematic flowchart of a watermark extraction step in Embodiment 1 of the present invention;

Fig. 3 is the principle block diagram of the electronic document digital watermark processing system in the second embodiment of the present invention;

4 is a schematic flowchart of a watermark embedding step in Embodiment 3 of the present invention;

5 is a schematic flowchart of a watermark extraction step in Embodiment 3 of the present invention;

Fig. 6a is a legend of the carrier image used in the simulation experiment in Embodiment 3 of the present invention;

Fig. 6b is the legend of the watermark image adopted in the simulation experiment in Embodiment 3 of the present invention;

Fig. 6c is the legend after embedding the watermark image in the simulation experiment in Embodiment 3 of the present invention;

6d is a legend of the watermark image extracted in the simulation experiment in Embodiment 3 of the present invention;

Fig. 7 is the legend of the electronic document image after embedding the watermark in Embodiment 3 of the present invention;

Fig. 8a is a legend of a carrier image of a Chinese document for performing a watermark processing test on an electronic document based on multiple features in the third embodiment of the present invention;

Figures 8b, 8c, and 8d are legends of watermark images to be embedded in the image shown in Figure 8a, respectively;

Fig. 8e is a legend after embedding 8b, 8c, 8d watermark images in Fig. 8a;

Fig. 9a is a legend of a carrier image of an English document for performing a watermarking test on an electronic document based on multiple features in the third embodiment of the present invention;

Figures 9b, 9c, and 9d are legends of watermark images to be embedded in the image shown in Figure 8a, respectively;

Fig. 9e is a legend after the watermark images 9b, 9c, and 9d are embedded in Fig. 9a;

The realization of the object of the present invention, its function and principle will be further described in the specific embodiments in conjunction with the accompanying drawings.

Detailed ways

Further description will be given below in conjunction with the accompanying drawings and specific embodiments.

Example 1:

As shown in FIG. 1 , this embodiment provides a digital watermark processing method for an electronic document, which mainly includes a watermark embedding step and a watermark extraction step. The watermark embedding step refers to the process of embedding the watermark image into the electronic document, and the watermark extraction step refers to the process of extracting the original watermark image from the digital watermark embedded electronic document, which is mainly used for identity authentication and so on.

Wherein, the above watermark embedding step mainly includes the following steps:

S11: Obtain user permissions;

S12: Collect watermark information, where the watermark information includes user information;

S13: upload the watermark information to the server;

S14: Receive the watermark information and store it in the database;

S15: Extract multiple features in the watermark information according to a predetermined algorithm, and perform feature fusion to generate a digital watermark;

S16: After converting the electronic document into an image, embed the digital watermark in the image, and then restore the image into an electronic document.

Specifically, obtaining the user authority in step S11 refers to a process of obtaining the user's operation authority by logging in the account and password of the registered user and after passing the verification. The user may be the creator of the electronic document, the copyright owner of the electronic document, or other interested parties.

Step S12 is used to collect watermark information. The watermark information in this embodiment refers to the original authentication or sign information to be embedded in the electronic document as a digital watermark, which may be the user's characteristic feature information, including the user's facial features and hand features. , one or more of fingerprint features and voice features.

After the collection is completed, upload the watermark information to the server.

Then, the server receives the watermark information and stores it in the database.

Meanwhile, for the watermark information, a plurality of features in the watermark information are extracted according to a predetermined algorithm, and feature fusion is performed to generate a digital watermark.

Finally, after converting the electronic document into an image, the above-mentioned digital watermark is embedded in the image, and then the image is restored into an electronic document.

The electronic documents referred to in this embodiment include, but are not limited to, office documents, pdf documents, and bitmap documents. In this way, after the watermark embedding operation on the electronic document is completed, the security and authenticity of the electronic document during transmission can be guaranteed.

Referring to Figure 2, the above-mentioned watermark extraction step mainly includes the following steps:

S21: Convert the digital watermark-embedded electronic document into an image;

S22: Process the image according to a predetermined algorithm to extract the digital watermark;

S23: Convert the digital watermark into a watermark image by inverse operation.

The watermark extraction step is mainly used to judge whether the digital watermark is complete or under attack, which is basically the reverse operation of the watermark embedding step, and will not be described in detail here.

In the process of transmission, digital images will inevitably be affected by some influences, such as compression, scanning, etc. This kind of attack is called unintentional attack, and some pirates deliberately destroy digital products, tamper with image content and even forge watermark information. Such attacks are called vicious assault. No matter what kind of attack, it will affect the detection of the extracted watermark, and the watermark embedding and extraction technology in this embodiment can effectively prevent the electronic document from being damaged or tampered with.

Embodiment 2:

As shown in FIG. 3 , this embodiment provides a digital watermark processing system 100 for electronic documents, which is mainly composed of two parts. One is a mobile client, which can be installed on mobile smart devices such as mobile phones, and its main functions are user registration, And upload the user's biometric picture as the feature information for the subsequent embedding of the watermark; one is the server side, which can be set up on the Internet remote or cloud server, and is mainly responsible for the preservation and management of the data and documents uploaded by the mobile client and the electronic documents. processing. The electronic document digital watermark processing system 100 can be divided into a watermark embedding unit 10 and a watermark extracting unit 20, wherein the watermark embedding unit 10 includes:

The authorization module 110, located in the mobile client, is used to obtain user rights;

An information collection module 120, located on the mobile client, is used to collect watermark information, where the watermark information includes user information;

an information uploading module 130, located on the mobile client, for uploading the watermark information to the server;

A database access module 140, located on the server side, is used to receive the watermark information and store it in the database of the server;

The watermark processing module 150, located on the server side, is used for extracting a plurality of features in the watermark information according to a predetermined algorithm, and performing feature fusion to generate a digital watermark;

The watermark embedding module 160, located on the server side, is used for converting the electronic document into an image, embedding the digital watermark in the image, and then restoring the image into an electronic document.

The watermark extraction units 20 are all located on the server side, and specifically include:

a conversion module 210, for converting the digital watermark-embedded electronic document into an image;

an extraction module 220, configured to process the image according to a predetermined algorithm to extract the digital watermark;

The restoration module 230 is used for converting the digital watermark into a watermark image by inverse operation.

The extracted watermark image can be fed back to the mobile client for users to view, save and authenticate.

This embodiment uses the user's biometric feature information as the watermark information, including but not limited to one or more of the user's facial feature, hand feature, fingerprint feature, and voice feature.

The mobile client is installed on an intelligent mobile device such as a mobile phone, and the information collection module includes a camera, a microphone, a fingerprint reader and/or a face recognizer of the device, which are used to collect facial features, hand features, and voices respectively. features and fingerprint features, etc.

This embodiment combines multiple biological features with digital watermarks, which can improve the security of the document system, protect the copyright of digital products, prove the authenticity of document content, track piracy, and determine whether the carrier has been tampered with. If the user needs to judge the authenticity of the content of the electronic document or track the copyright, the watermark can be extracted from the electronic document for further tampering detection and biometric authentication.

Embodiment three:

As shown in FIG. 4 , this embodiment further specifically provides a method for processing digital watermarking of an electronic document. According to statistics, there are a large number of white pixels in electronic document images. Based on this finding, this method selects an electronic document as a carrier image, and uses appropriate techniques to embed and extract the watermark, which includes two steps: one is the watermark embedding step, and the other is the reverse watermark extraction step.

The above watermark embedding steps specifically include:

S31: Convert the electronic document into a carrier image with a bit depth of 8;

S32: Select the white pixels to reconstruct the carrier image, that is, use the discrete wavelet algorithm to decompose the carrier image into four frequency domains: LL, HL, LH, HH, and select the HH frequency domain part as the watermark embedding area;

S33: Perform singular value decomposition on the watermark image;

S34: Embed the watermark image into the watermark embedding area, that is, replace the singular value of the watermark embedding area with the singular value of the watermark image, and use the inverse operation to obtain the reconstructed carrier sub-image embedded in the watermark;

S35: Restore the reconstructed carrier sub-image to the electronic document image according to the original pixel position.

The pixel value distribution of an image with a bit depth of 8 is generally concentrated in the middle and rear segments of the gray value. Through statistics on a large number of electronic documents, it is found that the pixel value is concentrated between 80 and 255, and there are a large number of white pixels. The situation only occurs with documentation.

The above watermark image contains the user's biometric information, including one or more of the user's facial features, hand features, fingerprint features and voice features.

In step S32, the HH frequency domain part is selected as the watermark embedding area, because this part mainly represents the edge and texture information of the original carrier image, and contains less subject information, so the watermark image is embedded in this frequency domain part to the carrier image. The quality impact is small, and the watermark embedding step of this embodiment adopts the following algorithm:

Assume that the electronic document image D is input, and the watermark image W _p×q . Extract the pixel point with a pixel value of 255 from the document picture, that is, the white pixel point, reconstruct it into a new carrier image A _m×n , and use the matrix P={p _ij 3 to record the extracted pixel point in the original carrier image. specific location.

The first-level discrete wavelet decomposition is performed on the reconstructed carrier image A _mxn , and four sub-images are generated, namely: LL, HL, LH, HH.

Singular value decomposition of the HH frequency domain subbands is performed using the following formula:

HH=U _h ×S _h ×V _h ^T :

Wherein, U _h is an m×m orthogonal matrix, V _h is an n×n orthogonal matrix, S _n =diag(σ ₁ ,σ ₂ ,...,σ _n ,), σ _i is a singular value of the image.

Then, perform singular value decomposition on the watermark image W:

WaterImage-U _s ×S _s ×V _s ^T ;

Next, the singular values S _h of the reconstructed carrier image are replaced by the singular values S _s of the watermark image.

Then, the inverse operation of singular value decomposition is used, as follows, to obtain the modified subgraph:

HH _new =U _h ×S _s ×V _s ^T ;

Finally, the inverse discrete wavelet transform is used to generate the image after the watermark is embedded, and then the image is restored according to the pixel position of the carrier image in the original document image, and the watermark-embedded electronic document picture can be obtained.

As we all know, the concept of wavelet transform was proposed by French engineer Morlet in 1974, discrete wavelet is to discretize continuous wavelet. Suppose the wavelet transform of any function x(t) is W(a, b), where a and b are scale factor and translation factor, respectively. Limit a and b in the wavelet basis function to some discrete points, that is, perform discrete sampling as shown in the following formula:

Then the calculation of wavelet Ψ _{a, b} (t) is as follows:

The discrete wavelet transform (DWT) is defined as Equation 2-3:

DWT(m,n)=∫ _R f(t)Ψ _m,n (t)dt:

Discrete wavelet transform is a wavelet transform based on finite time domain and transform frequency, which is a frequency domain technique. The image is stored in the form of a two-dimensional matrix, so the wavelet transform of the image is based on the two-dimensional discrete wavelet transform. This algorithm applies the one-dimensional wavelet transform to the rows and columns of the two-dimensional signals such as images, etc. Two-dimensional wavelet transform can perform effective time-frequency decomposition on images. Discrete wavelet transform firstly transforms the carrier image to frequency domain, and then modifies the frequency coefficient of the original carrier image according to the frequency coefficient of the transformed watermark image, so as to obtain a robust watermark-embedded image.

Discrete wavelet transforms decompose images hierarchically, providing spatial and frequency image descriptions. It decomposes the image in three basic spatial directions, namely horizontal, vertical and diagonal, resulting in four different components, namely LL band, LH band, HL band and HH band. The first letter here refers to low-pass frequency operation or high-pass frequency operation on the rows of the carrier image, and the second letter refers to applying the filter to the columns of the carrier image. The LL band represents the detail information obtained by low-pass filtering in the horizontal and vertical directions, the LH band represents the detail information obtained after low-pass filtering in the horizontal direction and the high-pass filtering in the vertical direction, and the HL band represents the high-pass in the horizontal direction and the low-pass filtering in the vertical direction. Detail information, the HH band represents the detail information obtained after high-pass filtering in the horizontal and vertical directions. The LL band is the lowest resolution level and consists of the approximate part of the carrier image, which concentrates most of its information and depicts the main features. The other three bands maintain the edge detail information of the carrier image in each direction.

For the second-level decomposition, any sub-band can be further decomposed into four levels. The higher the decomposition level, the better the robustness of the watermark-embedded image. At each decomposition level, the wavelet coefficients of the LL band are orders of magnitude larger than the other three bands. The human visual system is more sensitive to the low frequency part, that is, the LL frequency band, so the watermark information is preferably embedded in the other three frequency bands to better preserve the quality of the original image.

After the two-dimensional wavelet transform, the image is decomposed into 4 sub-block regions whose size is 1/4 of the original size, respectively including the wavelet coefficients of the corresponding frequency bands, which is equivalent to sampling at intervals in the horizontal and vertical directions. The algorithm of this embodiment selects the HH sub-band (high-frequency sub-band) as the watermark embedding area, because this frequency band represents the edge and texture information of the image, and the quality of the original image is less affected after the watermark is embedded.

Singular value decomposition belongs to linear algebra and is an asymmetric orthogonal transformation, which is based on eigenvectors and is widely used in digital image processing. The singular value does not change significantly when the image is disturbed, and the stability is good. The singular value represents the essential characteristics of an image. After the watermark is embedded, the singular value changes very little when the image is subjected to a simple attack. It is precisely because it does not represent the visual characteristics of the image that the singular value decomposition has better concealment and resistance to geometric attacks, and can well make up for the defect that the discrete wavelet transform cannot well resist geometric attacks, so the singular value of the image. Value decomposition and discrete wavelet transform can be well combined to improve the robustness of digital watermarking and the execution rate of digital watermarking algorithms, and also make more choices of embedded coefficients.

The basic idea of image singular value decomposition: if a digital image is represented by a matrix A, set the matrix ^{A∈Rm ×n} , rank(A)=r, r≤n, then the singular value decomposition of matrix A is defined as follows shown:

where U=[u ₁ , u ₂ ,..., _um ,]∈R ^m×m and V=[v ₁ ,v ₂ ,...,v _n ,]∈R ^n×n are orthogonal matrices whose columns The vectors are _ui and v _i respectively; U and V are called the left singular matrix and right singular matrix of matrix A respectively; D=diag(σ ₁ ,σ ₂ ,...,σ _m ,) is a diagonal matrix, σ _i ( i=1, 2, ..., m) is called the singular value of matrix A, here is the positive square root of the eigenvalue of ^{AAT or A T A} , and satisfies σ ₁ ≥σ ₂ ≥...≥σ _r >σ _{r+ 1} =...=σ _m =0. The geometric meaning of singular value decomposition is that for a matrix of any size, it is necessary to find a set of unit vectors that are orthogonal to each other, so that the matrix interacts with these two vectors to obtain two new vectors, and it is necessary to maintain these two new vectors. The resulting vectors are also orthogonal.

The singular value of the image characterizes the intrinsic properties of the image and has good stability. When the image has a small change, its singular value will not change greatly, and the singular value of the matrix has transposition invariance and rotation invariance. , these characteristics are of great significance to realize the robustness of digital watermarking.

As shown in Figure 5, the watermark extraction step is basically the reverse operation of the watermark embedding step, as follows:

S41: extracting the pixel points of the watermark-embedded electronic document image to obtain a watermark-embedded carrier image;

S42: Perform discrete wavelet transform on the watermark-embedded carrier image, and select the HH frequency domain part as the watermark extraction area;

S43: Perform singular value decomposition on the watermark extraction area, and extract singular values of the watermark extraction area;

S44 : use the inverse operation to restore the watermark image by using the obtained singular values.

The algorithm of this embodiment is simulated in the Matlab R2012b environment. The experimental carrier image is a 512×512 electronic document image, as shown in Figure 6a; the watermark image is a 256×256 palm grayscale image, as shown in Figure 6b ; the image after embedding the watermark is shown in Figure 6c; the extracted watermark image is shown in Figure 6d. The following will evaluate the invisibility and robustness of the algorithm through experiments, and compare it with other watermarking algorithms.

The image information embedded in the watermark is transmitted in the channel. In this process, it is generally subject to various types of attacks, and the watermark information embedded in it will also be damaged to varying degrees. Some malicious attackers will steal the watermark information, and some even want to tamper and forge data, so that the digital watermarking algorithm needs to have security against attacks. Common geometric attacks mainly include compression, noise, filtering, rotation scaling, etc. In order to determine the attack resistance of the watermarking algorithm, it needs to be tested according to some criteria. The evaluation index of the digital image watermarking algorithm is mainly the invisibility and stability of the watermark. After inspection, the method of this embodiment has higher stability and better watermark invisibility for the electronic document embedded with the watermark.

The following is the evaluation and verification process of watermark invisibility.

The invisibility of the watermark is also called imperceptibility. Intuitively, the human eye cannot feel the difference between the original image and the image after the watermark is embedded. The better the invisibility, the harder it is for people to find the hidden watermark. The security of the watermark information is better guaranteed, and the good invisibility also makes the quality of the original image not greatly affected.

Generally, peak signal-to-noise ratio (PSNR) and structural similarity (SSIM) are used as evaluation indicators.

The peak signal-to-noise ratio is used to describe the quality change of the image before and after watermark embedding. The watermark embedding system is regarded as a communication system. The original carrier image represents the signal to be transmitted, and the embedded watermark information represents the noise loaded on the original signal. The noise ratio can also be understood as the ratio of the maximum value of the transmission signal to the noise, and its calculation formula is shown in the following formula.

Among them, MSE represents mean square error, and the calculation formula is:

Among them, I _M×N represents the original carrier image, and I′ _M×N represents the image after embedding the watermark. The mean square error is used to calculate the difference between the two images, and its size can indicate the degree of fluctuation of the image difference.

Under normal circumstances, when the value of the peak signal-to-noise ratio is not less than 36dB, it is considered acceptable for the watermarking algorithm to damage the quality of the original carrier image. The less distortion, the higher the quality of the image after embedding the watermark.

Another metric to measure the similarity of two images is Structural Similarity (SSIM), which is calculated as follows:

是x的方差，

是y的方差，σ_xy是x和y的标准差。c₁＝(k₁L)²，c₂＝(k₂L)²是用来维持稳定的常数。L是像素值的动态范围，经过实验总结，令k₁＝0.01，k₂＝0.03。结构相似性值越大，表示图像失真越小。where u _x is the mean of x, u _y is the mean of y,

is the variance of x,

is the variance of y, and σ _xy is the standard deviation of x and y. c ₁ =(k ₁ L) ² and c ₂ =(k ₂ L) ² are constants for maintaining stability. L is the dynamic range of the pixel value, and after summarizing the experiment, let k ₁ =0.01, k ₂ =0.03. The larger the structural similarity value, the smaller the image distortion.

Images are highly structured and there are strong correlations between pixels, especially when they are spatially similar. Most error-sensitivity-based quality assessment methods, such as peak signal-to-noise ratio and mean square error, decompose image signals based on linear transformations, which do not involve image-specific correlations. Therefore, an evaluation index that measures the similarity of the structure can be selected.

In this experiment, three different electronic document images are used as carrier images, and they are all embedded with the same watermark image as shown in Figure 6b for experimental comparison. The experimental results are shown in Figure 7. Table 1-1 shows the three embedded watermark images. Peak signal-to-noise ratio and structural correlation coefficient, from the data in the table, it can be seen that the algorithm in this paper has good invisibility and data fidelity.

Table 1-1 PSNR value and SSIM value of watermarking algorithm

In this experiment, the watermark image shown in Figure 6a is embedded in Figure 6b, and its peak signal-to-noise ratio and structural similarity are calculated, and compared with other algorithms. Table 1-2 shows the comparative test results. It can be seen from the experimental data that the peak signal-to-noise ratio of the algorithm proposed in this paper is 58.81, which is higher than that of other algorithms, and the structural correlation coefficient is the maximum value, so compared with other algorithms, it has a higher peak signal-to-noise ratio. Good concealment.

Table 1-2 PSNR and SSIM comparison test

The following is an experiment for watermarking electronic documents based on multiple features.

Select two electronic document pictures as carrier images, one is full Chinese and the other is full English. These two documents have the most fonts in Chinese and English languages, that is, the case where the white pixels are relatively few, so these two carrier images It can fully explain the capacity of the watermarking algorithm in this paper. The experiments in this section embed three different watermark images in the two carrier images respectively. Three palm images are embedded in the Chinese electronic document, and two face images and one palm image are embedded in the English electronic document image.

The size of the carrier image used in the experiment is equal to 1024×1024, the size of the embedded watermark image is 128×128, and the watermark image is collected by an Android mobile phone. The experimental results are shown in Figure 8e and Figure 9e. It can be seen from the experimental results that embedding multiple biometric images will not affect the readability of the original carrier image, and the extracted watermark image is also highly similar to the original feature image. It also provides conditions for biometric recognition and multi-biometric fusion.

Embedding experiments for the above two different carrier images and different watermark images, evaluate the fidelity and invisibility of the images after embedding the watermark, and test the peak signal-to-noise ratio (PSNR) and structural similarity (SSIM) after embedding the watermark. , from the data shown in Table 2-1, it can be seen that the PSNR value of the algorithm in this paper is relatively high, so embedding multiple watermark images will not affect the readability of the original carrier.

Table 2-1 Experimental results of embedding multiple watermark images

In the experiment, watermark extraction is performed on the above two watermarked images, and the normalized correlation coefficient between the extracted watermark and the original watermark image is calculated to judge whether the watermark image is distorted. The experimental data is shown in Table 2-2. Whether it is an English document or a Chinese document image, the three watermark images extracted from it maintain a high correlation with the original image, which also shows that the algorithm of this embodiment is applied to multiple watermarks It is also feasible and robust when embedding.

Table 2-2 Extract the NCC value of the watermark image

To sum up, the present invention proposes a digital watermark processing method and system for electronic documents according to the characteristics of the pixel distribution of electronic documents. The electronic documents have a large number of white pixels, which are interspersed between the lines. It is a feature that does not exist in other natural images. Through these pixels, a new carrier image is reconstructed, the watermark is embedded in it, and then restored to the original pixel position. This method of reconstruction and restoration also realizes the embedding of the watermark. However, the watermark embedding method of the present invention operates in the image frequency domain, and combines discrete wavelet transform and singular value decomposition technology to achieve watermark embedding. The experimental results show that the method has good robustness. In addition, by analyzing the capacity of white pixels in the electronic document, the present invention can realize 2 to 3 kinds of biological features as watermark images. It is proved by experiments that the embedding of multiple watermarks does not have a great impact on the quality of the carrier image, and the watermarking algorithm is still It has better invisibility and robustness.

The technical features of the above-described embodiments can be combined arbitrarily. For the sake of brevity, all possible combinations of the technical features in the above-described embodiments are not described. However, as long as there is no contradiction between the combinations of these technical features, All should be regarded as the scope described in this specification.

The above-mentioned embodiments only represent several embodiments of the present invention, and the descriptions thereof are specific and detailed, but should not be construed as a limitation on the scope of the patent of the present invention. It should be pointed out that for those skilled in the art, without departing from the concept of the present invention, several modifications and improvements can be made, which all belong to the protection scope of the present invention.

Claims (10)

1. an electronic document digital watermark processing method, is characterized in that, comprises watermark embedding step, and described watermark embedding step specifically comprises: Obtain user permissions; Collect watermark information, the watermark information includes user information; uploading the watermark information to the server; Receive the watermark information and store it in the database; Extract multiple features in the watermark information according to a predetermined algorithm, and perform feature fusion to generate a digital watermark; After the electronic document is converted into an image, the digital watermark is embedded in the image, and the image is restored into an electronic document. 2. The electronic document digital watermark processing method according to claim 1, further comprising a watermark extraction step, the watermark extraction step specifically comprising: Convert digitally watermarked electronic documents into images; Process the image according to a predetermined algorithm to extract the digital watermark; Using an inverse operation, the digital watermark is converted into a watermarked image. 3. electronic document digital watermark processing method as claimed in claim 1 or 2, is characterized in that, described watermark information is user's biometric information, including user's facial feature, hand feature, fingerprint feature and voice feature. one or more of them. 4. An electronic document digital watermark processing system, characterized in that it comprises a watermark embedding unit, and the watermark embedding unit specifically comprises: Authorization module, used to obtain user permissions; an information collection module for collecting watermark information, where the watermark information includes user information; an information uploading module for uploading the watermark information to the server; a database access module for receiving the watermark information and storing it in the database of the server; a watermark processing module, used for extracting multiple features in the watermark information according to a predetermined algorithm, and performing feature fusion to generate a digital watermark; The watermark embedding module is used to embed the digital watermark in the image after converting the electronic document into an image, and then restore the image into an electronic document. 5. The electronic document digital watermark processing system of claim 4, further comprising a watermark extraction unit, the watermark extraction unit specifically comprising: The conversion module is used to convert the digital watermark-embedded electronic document into an image; an extraction module, configured to process the image according to a predetermined algorithm to extract the digital watermark; The restoration module is used for converting the digital watermark into a watermark image by inverse operation. 6. The electronic document digital watermark processing system according to claim 5, wherein the electronic document digital watermark processing system is composed of a mobile client and a server, wherein the mobile client comprises the authorization module, An information collection module and an information upload module, the server side includes the database, a database access module, a watermark processing module, a watermark embedding module, a watermark decryption module, a search module and an extraction module. 7. The electronic document digital watermark processing system according to claim 6, wherein the watermark information is the biometric information of the user, including one of the user's facial feature, hand feature, fingerprint feature and voice feature. species or several; The mobile client is installed on a mobile phone, and the information collection module includes a camera, a microphone, a fingerprint identifier and/or a face identifier of the mobile phone. 8. A method for processing digital watermarking of electronic documents, comprising a watermark embedding step of embedding a watermark image in an electronic document, the watermark embedding step specifically comprising: Convert electronic documents into carrier images with a bit depth of 8; Select the white pixels to reconstruct the carrier image, that is, use the discrete wavelet algorithm to decompose the carrier image into four frequency domains: LL, HL, LH, HH, and select the HH frequency domain part as the watermark embedding area; performing singular value decomposition on the watermark image; Embed the watermark image into the watermark embedding area, that is, replace the singular value of the watermark embedding area with the singular value of the watermark image, and use the inverse operation to obtain the reconstructed carrier sub-image embedded in the watermark; The reconstructed carrier sub-image is restored to the electronic document image according to the original pixel position. 9. The electronic document digital watermark processing method according to claim 8, further comprising a watermark extraction step, the watermark extraction step specifically comprising: Extracting the pixel points of the watermark-embedded electronic document image to obtain the watermark-embedded carrier image; Discrete wavelet transform is performed on the carrier image embedded in the watermark, and the HH frequency domain part is selected as the watermark extraction area; Perform singular value decomposition on the watermark extraction area, and extract the singular values of the watermark extraction area; Using the inverse operation, the watermark image is restored by using the obtained singular values. 10. The electronic document digital watermark processing method according to claim 8 or 9, wherein the watermark image comprises the user's biometric information, including the user's facial feature, hand feature, fingerprint feature and voice feature. one or more of them.

Images