CN112800395B - A Multi-Image Copyright Authentication and Verification Method Based on Zero Watermark Technology - Google Patents

Abstract

The invention discloses a multiple image copyright authentication method based on zero watermark technology, comprising the following steps: performing image processing on multiple images to be protected, and fusing them into one image; determining the effective area of the fused image, and Features are extracted in the region to obtain a feature image; the feature image and the logo image are respectively scrambled and encrypted, and the scrambled feature image and the scrambled logo image are XORed to obtain a zero-watermark image. A method for verifying the copyright of multiple images based on zero-watermark technology, comprising the following steps: performing an anti-XOR operation on a scrambled feature image and a zero-watermark image to obtain a scrambled logo image, and then performing scrambled decryption to obtain An identification image is verified and compared with the original identification image. The method protects the copyrights of multiple images more efficiently and safely at the same time, effectively reduces the time and storage costs in the copyright protection process, and has great practical value.

Description

A copyright authentication and verification method for multiple images based on zero watermark technology

Technical Field

The invention belongs to the technical field of digital multimedia anti-counterfeiting and information security protection, and specifically relates to a multi-image copyright authentication and verification method based on zero watermark technology.

Background Art

With the rapid development of the Internet, computers and mobile communication technologies, people can upload or download multimedia information (images, videos, texts and audio, etc.) more conveniently and quickly. However, while the rapid development of technology brings convenience to people, it also brings information security issues such as easy duplication and theft of multimedia information. In recent years, the rapid development of digital watermarking technology has solved the problem of multimedia information security protection to a certain extent. Traditional embedded watermarking technology is the most widely used image copyright protection technology in recent years. However, as the watermark information is embedded in the original image, the image data will inevitably be modified, thereby affecting the image quality.

At present, most zero-watermark algorithms are aimed at copyright protection of a single image. When copyright protection is performed on a large-scale image set, repeated operations will not only consume a lot of time, but also occupy a lot of storage space for copyright evidence. Therefore, how to reduce the time and storage costs in copyright protection will become a research hotspot in the future. In 2016, Shao Zhuhong et al. proposed a robust zero-watermark scheme based on orthogonal Fourier Mellin moments and chaotic mapping, which can realize copyright protection of two images at the same time. First, the two images are respectively regarded as the real and imaginary parts of complex numbers, so that the "two-channel structure" is merged into a "single-channel structure", and then the characteristic invariants obtained by the orthogonal Fourier Mellin moments are calculated, and the binary feature image is constructed using it. Finally, the watermark image and the feature image are scrambled together using chaotic mapping to generate a verification image. This scheme has certain robustness and security. However, this method maps the image to the complex domain, so it can only process two images at the same time, and the algorithm may be difficult to deal with malicious attacks such as translation, mirroring and joint attacks. In 2019, Xia Zhiqiu et al. proposed a scheme for copyright protection of three CT images simultaneously based on the quaternion pole harmonic Fourier moment. The scheme first maps the three CT images into three imaginary parts of pure quaternions, thereby merging the "three-channel structure" into a "single-channel structure", and then calculates its quaternion pole harmonic Fourier moment. Then, the amplitude of the quaternion pole harmonic Fourier moment is used to construct a binary feature image. Finally, the feature image is chaotically scrambled and XORed with the watermark image to generate a key image.

In summary, although the current multi-image zero-watermark algorithms can map multiple images to the complex domain, thereby achieving copyright protection for two or three images at the same time, these algorithms lack flexibility, that is, they cannot flexibly determine a reasonable image copyright protection scheme and the number of images to be protected at one time according to the number of large-scale image sets, and cannot provide copyright protection for all images in a large-scale image set.

Summary of the invention

The technical problem to be solved by the present invention is to provide a copyright authentication and verification method for multiple images based on zero watermark technology in view of the deficiencies of the above-mentioned prior art.

In order to achieve the above technical objectives, the following technical solutions are adopted:

A method for copyright authentication of multiple images based on zero watermark technology comprises the following steps:

Step S1: performing image processing on a plurality of images to be protected and fusing them into a fused image;

Step S2: determining the effective area of the fused image, and extracting features in the effective area to obtain a feature image;

Step S3: The feature image and the identification image are scrambled and encrypted respectively, and the scrambled feature image and the scrambled identification image are XOR-ed to obtain a zero-watermark image.

To optimize the above technical solutions, the specific measures taken also include:

Furthermore, step S1 is specifically as follows:

S11: performing shear normalization in the X-axis direction, shear normalization in the Y-axis direction, scaling normalization, and translation normalization on the multiple images using an image normalization method to obtain corresponding standard images;

S12: Multiple standard images are fused into a fused image using the grayscale weighted average image fusion method. The fusion formula is:

Im _F ＝α ₁ *Im ₁ +α ₂ *Im ₂ +…+α _i *Im _i

α ₁ =α ₂ =…=α _i , α ₁ +α ₂ +…+α _i =1, i∈N ₊

Wherein: Im ₁ , Im ₂ , ..., Im _i represent multiple images to be fused, Im _F represents an image obtained after the multiple images are fused, and i represents the number of the multiple images to be fused.

Furthermore, step S2 is specifically as follows:

S21: extracting a valid area of a fixed size according to the centroid of the fused image;

S22: Perform lifting wavelet transform on the effective area to obtain a low-frequency component matrix of the image signal in the effective area, decompose the low-frequency component matrix into multiple sub-matrix blocks A of fixed size, and then perform QR decomposition on each sub-matrix block A to obtain an orthogonal matrix Q and an upper triangular matrix R.

The decomposition formula of the low-frequency component matrix A is: A = QR

Where: Q is an orthogonal matrix, R is an upper triangular matrix;

S23: Calculate the mean of the matrix elements composed of the 2-norm of the first row vector of the upper triangular matrix R. If the element value is greater than or equal to the mean, take 1, otherwise take 0, so as to obtain a binary feature image.

Furthermore, the scrambling encryption process in step S3 is: according to the key K, the feature image and the identification image are scrambled and encrypted and decrypted using Cat mapping.

The encryption formula is:

Where: ( _xn , _yn ) represents the coordinates of the original grayscale image pixel, ( _xn+1 , yn ₊₁ ) represents the pixel coordinates after transformation, a and b are scrambling parameters, n represents the number of current transformations, mod() is the modulus operation, and N is the order of the image. The scrambling parameters and the number of transformations constitute the key K.

A multi-image copyright verification method based on zero watermark technology, according to the key K, Cat mapping is used to scramble and encrypt the feature image to obtain a scrambled feature image, the scrambled feature image is de-XORed with the zero watermark image to obtain a scrambled identification image, and then the scrambled decryption is performed according to the key K to obtain the identification image, and the identification image is compared with the original identification image to obtain a similarity coefficient. When the similarity coefficient is greater than a set threshold, it is considered that the copyright verification is correct, otherwise it fails;

The decryption formula of the scrambled logo image is:

Beneficial effects of the present invention:

The present invention discloses a method for copyright authentication and verification of multiple images based on zero watermark technology. The method can fuse multiple images into one image by using a grayscale weighted average image fusion method. Therefore, only one fused image needs to be subjected to feature extraction and other operations, thereby effectively reducing the time and storage cost caused by repeated operations on a single image in copyright protection. At the same time, according to the number of images in a large-scale image collection, a reasonable protection scheme and the number of images to be protected at one time are determined, so that the protection method has a certain flexibility, thereby realizing copyright protection of all images in the image collection. This effective effect is obvious.

Secondly, the present invention uses image normalization technology to convert multiple images into standard images, and uses LWT-QR decomposition to perform feature extraction. Compared with the zero watermark algorithm of a single image, the present invention has higher robustness in dealing with image attacks. At the same time, the use of multiple image fusion method and LWT-QR decomposition for feature extraction can effectively save algorithm running time, and can protect the copyright of multiple images more efficiently and safely, effectively reducing the time and storage costs in the copyright protection process, and has great practical value.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG1 is a flow chart of the copyright authentication method of the present invention.

FIG. 2 is a schematic diagram of the copyright authentication algorithm flow of the present invention.

FIG3 is a schematic diagram of the copyright verification algorithm flow of the present invention.

FIG. 4 is a diagram comparing the robustness of different solutions of the present invention.

FIG5 is a comparison chart of the running time of different schemes of the present invention.

FIG. 6 is a line graph comparing the robustness of different solutions of the present invention.

DETAILED DESCRIPTION

The embodiments of the present invention are further described in detail below in conjunction with the accompanying drawings.

This method first normalizes multiple images to corresponding standard images, then fuses multiple standard images into one image using the grayscale weighted average image fusion method, then uses LWT-QR decomposition to extract features from the effective area of the fused image, and finally uses Cat mapping to scramble the feature image and the identification image, and performs an XOR operation on the scrambled feature image and the identification image to obtain the key to copyright authentication and verification - the zero-watermark image. This solves the time and storage cost problems caused by repeated operations of the zero-watermark algorithm for a single image when protecting the copyright of a large-scale image set, and also solves the problem that the existing zero-watermark algorithm for multiple images lacks flexibility and cannot protect the copyright of all images in the image set.

The method includes two stages: copyright authentication and copyright verification. The two stages are respectively described in conjunction with FIG1 and FIG2 .

As shown in FIG1 , the present invention is a method for copyright authentication of multiple images based on zero watermark technology, comprising the following steps:

Step S1: Process multiple images to be protected and fuse them into one image.

Step S1 is specifically as follows:

S11: performing shear normalization in the X-axis direction, shear normalization in the Y-axis direction, scaling normalization, and translation normalization on the multiple images using an image normalization method to obtain corresponding standard images.

S12: Multiple standard images are fused into a fused image using the grayscale weighted average image fusion method. The fusion formula is:

Im _F ＝α ₁ *Im ₁ +α ₂ *Im ₂ +…+α _i *Im _i

α ₁ =α ₂ =…=α _i , α ₁ +α ₂ +…+α _i =1, i∈N ₊

Wherein: Im ₁ , Im ₂ , ..., Im _i represent multiple images to be fused, Im _F represents an image obtained after the multiple images are fused, and i represents the number of the multiple images to be fused.

Step S2: determining the effective area of the fused image, and extracting features in the effective area to obtain a feature image;

Step S2 is specifically as follows:

S21: Extracting a valid area of a fixed size according to the centroid of the fused image.

S22: Perform lifting wavelet transform on the effective area to obtain a low-frequency component matrix of the image signal in the effective area, decompose the low-frequency component matrix into multiple sub-matrix blocks A of fixed size, and then perform QR decomposition on each sub-matrix block A to obtain an orthogonal matrix Q and an upper triangular matrix R.

The decomposition formula of the low-frequency component matrix A is: A = QR

Among them: Q is an orthogonal matrix, and R is an upper triangular matrix.

S23: Calculate the mean of the matrix elements composed of the 2-norm of the first row vector of the upper triangular matrix R. If the element value is greater than or equal to the mean, take 1, otherwise take 0, so as to obtain a binary feature image.

Step S3: The feature image and the identification image are scrambled and encrypted respectively, and the scrambled feature image and the scrambled identification image are XOR-ed to obtain a zero-watermark image.

The scrambling encryption process is: according to the key K, the feature image and the identification image are scrambled and encrypted using Cat mapping.

The encryption formula is:

Where: ( _xn , _yn ) represents the coordinates of the original grayscale image pixel, ( _xn+1 , yn ₊₁ ) represents the pixel coordinates after transformation, a and b are scrambling parameters, n represents the number of current transformations, mod() is the modulus operation, and N is the order of the image. The scrambling parameters and the number of transformations constitute the key K.

Embodiment 1

第i个子块记为A_i,i＝1,2,...,K，令

First, four different original images of size M×M are normalized respectively to convert them into standard images. Then, the grayscale weighted average image fusion method is used to fuse the four standard images to obtain a fused image containing the information of each image. Then, a fixed-size N×N effective area is extracted based on the indeformed center of the fused image. Then, a level l lifting wavelet transform is performed on the effective area to obtain a low-frequency component matrix, which is divided into sub-matrix blocks A of size n×n. The total number of sub-matrix blocks A is

The i-th sub-block is denoted as A _i , i=1,2,...,K, let

Then, QR decomposition is performed on each sub-matrix block _Ai , and the 2-norm of the first row vector of the R matrix of each sub-matrix block _Ai is calculated to form a matrix B of size k×k. Subsequently, the mean ave of the elements in the matrix B is calculated, and ave is used as the basis for generating the feature image. If the value of the element in the matrix B is greater than or equal to the mean ave, it is taken as 1, and the value of the element in the matrix B is less than the mean ave, it is taken as 0, and the corresponding feature image F is generated.

从而得到零水印图像ω，同时，将零水印图像ω保存至第三方可信机构，作为版权验证的依据。Finally, the Cat mapping is used to scramble the feature image F and the identification image W, where the scrambling parameter is saved by the copyright owner as the key K for copyright verification. Then, the scrambled feature image F' and the scrambled identification image W' are XORed. The formula for the XOR operation is:

Thus, a zero-watermark image ω is obtained. At the same time, the zero-watermark image ω is saved to a third-party trusted organization as a basis for copyright verification.

A copyright verification method for multiple images based on zero watermark technology, specifically: according to a key K, a feature image is encrypted by using Cat mapping to obtain a scrambled feature image, the scrambled feature image is de-XORed with a zero watermark image to obtain a scrambled identification image, and then the scrambled decryption is performed according to the key K to obtain an identification image, and the identification image is compared with the original identification image to obtain a similarity coefficient. When the similarity coefficient is greater than a set threshold, it is considered that the copyright verification is correct, otherwise it fails.

The decryption formula of the scrambled logo image is:

Embodiment 2

In the copyright verification process, the copyright owner needs to provide the key K and the identification image to prove his ownership of the image. As shown in Figures 1 and 2, the feature extraction steps of verification are the same as those of authentication, so we will not go into details here, and only introduce the different parts in detail.

According to the key K, the generated feature image F is scrambled using Cat mapping, and the scrambled feature image F' is subjected to an anti-XOR operation with the zero watermark image ω stored by a trusted third-party organization to obtain the scrambled identification image W'. The anti-XOR operation formula is:

Finally, the scrambled logo image W' is descrambled using the key K to restore the logo image W. The restored logo image is compared with the logo image saved by the copyright owner to obtain a similarity coefficient. If it is greater than the set threshold, it is considered copyright verification, otherwise it fails.

Figures 4 and 6 show the robustness of the proposed scheme and compare the robustness with the three schemes. Scheme 1 is a copyright protection scheme based on discrete cosine transform. This scheme uses normalization processing in both watermark embedding and extraction processes, and can resist certain malicious signal processing and geometric attacks. Scheme 2 uses image normalization technology to map the image to a geometrically invariant space. Then, the important area is extracted from the normalized image. Finally, the low-frequency discrete cosine transform coefficients of the important area and the watermark image are XORed to construct copyright information. Scheme 3 is a strong robust zero watermark algorithm based on non-subsampled Contourlet transform and image normalization. This scheme uses image normalization technology based on invariant moments to convert the original image into a standard image. Then, the effective area is subjected to block non-subsampled Contourlet transform. The low-frequency coefficient subbands obtained by the transform are subjected to block singular value decomposition, and a zero watermark is generated according to the parity of the highest bit of the maximum singular value of each block. Compared with the other three schemes, the method of this scheme has higher robustness in dealing with image attacks, as shown in Figures 4 and 6. At the same time, the use of multiple image fusion methods and LWT-QR decomposition for feature extraction can effectively save the algorithm running time, as shown in Figure 5, so it has a lower time cost. From the above charts, it can be seen that the method proposed in this scheme is an optimal solution.

The above are only preferred embodiments of the present invention. The protection scope of the present invention is not limited to the above embodiments. All technical solutions under the concept of the present invention belong to the protection scope of the present invention. It should be pointed out that for ordinary technicians in this technical field, some improvements and modifications without departing from the principle of the present invention should be regarded as the protection scope of the present invention.

Claims (2)

1. A method for copyright authentication of multiple images based on zero watermark technology, characterized by comprising the following steps: Step S1: Process multiple images to be protected and fuse them into a fused image; specifically: S11: performing shear normalization in the X-axis direction, shear normalization in the Y-axis direction, scaling normalization, and translation normalization on the multiple images using an image normalization method to obtain corresponding standard images; S12: Multiple standard images are fused into a fused image using the grayscale weighted average image fusion method. The fusion formula is: Im _F ＝α ₁ *Im ₁ +α ₂ *Im ₂ +…+α _i *Im _i α ₁ =α ₂ =…=α _i , α ₁ +α ₂ +…+α _i =1, i∈N ₊ Where: Im ₁ ,Im ₂ ,...,Im _i represent multiple images to be fused, Im _F represents the image obtained after the fusion of multiple images, and i represents the number of images to be fused; Step S2: Determine the effective area of the fused image, and extract features in the effective area to obtain a feature image; specifically: S21: extracting a valid area of a fixed size according to the centroid of the fused image; S22: performing lifting wavelet transform on the effective area to obtain a low-frequency component matrix of the image signal in the effective area, decomposing the low-frequency component matrix into a plurality of sub-matrix blocks A of a fixed size, and then performing QR decomposition on each sub-matrix block A to obtain an orthogonal matrix Q and an upper triangular matrix R; The decomposition formula of the low-frequency component matrix A is: A = QR Among them: the size of the orthogonal matrix Q is m×m, and the size of the upper triangular matrix R is m×n; S23: Calculate the mean of the matrix elements composed of the 2-norm of the first row vector of the upper triangular matrix R. If the element value is greater than or equal to the mean, it is set to 1, otherwise, it is set to 0, thereby obtaining a binary feature image; Step S3: The feature image and the identification image are scrambled and encrypted respectively, and the scrambled feature image and the scrambled identification image are XORed to obtain a zero watermark image. The scrambled encryption process is specifically as follows: According to the key K, the feature image and the identification image are scrambled and encrypted using Cat mapping. The encryption formula is:

Where: ( _xn , _yn ) represents the coordinates of the original grayscale image pixels, ( _xn+1 , yn ₊₁ ) represents the pixel coordinates after transformation, a and b are scrambling parameters, n represents the number of current transformations, mod() is the modular operation, N is the order of the image, and the scrambling parameters and the number of transformations constitute the key K. 2. A method for copyright verification of multiple images based on zero watermark technology, characterized in that: according to a key K, a feature image is encrypted by using Cat mapping to obtain a scrambled feature image, the scrambled feature image is subjected to an inverse XOR operation with the zero watermark image to obtain a scrambled identification image, and then the scrambled decryption is performed according to the key K to obtain an identification image, and the identification image is compared with the original identification image for similarity to obtain a similarity coefficient. When the similarity coefficient is greater than a set threshold, the copyright verification is considered to be correct, otherwise it fails; The zero-watermark image is obtained by the multiple-image copyright authentication method based on zero-watermark technology according to claim 1; The decryption formula of the scrambled logo image is:

Images