Color image robust reversible watermarking method based on quaternion Zernike moment

Abstract

The invention provides a color image robust reversible watermarking method based on a quaternion Zernike moment, and belongs to the field of image security. The problems of low efficiency and insufficient robustness of an existing method for embedding a robust reversible watermark algorithm into a color image are solved. The method comprises the following steps: firstly, embedding a robust watermark into a quaternion Zernike moment coefficient of a color image; then, according to the watermark embedded with the robustness, an intermediate image is generated, and the intermediate image is close to the original color image; and converting the difference value between the intermediate image and the original color image into a bit sequence, and embedding the bit sequence into the image added with the robust reversible watermark by adopting a reversible information hiding mode to finally obtain the image added with the robust reversible watermark. When the watermark is extracted, the original color image can be recovered from the image added with the robust reversible watermark, the method has strong robustness in the aspects of salt and pepper noise, Gaussian noise, JPEG compression and image blurring, the robust reversible watermark is embedded into the color image, and the method can be applied to copyright protection of color medical images.

Description

Technical Field

[0001] The invention provides a color image robust reversible watermark method based on quaternion Zernike moments, belonging to the technical fields of information hiding and image security. Background Art

[0002] With the development of multimedia technology, digital images can be used to declare copyright by adding robust watermarks during network transmission and use. However, medical images have high requirements for image quality during use, and traditional methods of adding robust watermarks will distort the original color images, increasing the risk of misdiagnosis of medical images.

[0003] In order to solve this problem, robust reversible watermarking technology becomes particularly important. On the one hand, robust reversible watermarking technology can embed a robust watermark in an image to declare the copyright of the image, and on the other hand, it can restore the original color image from the image embedded with the robust watermark, ensuring the reversibility of the image embedded with the robust watermark. However, traditional robust reversible watermarking technology usually processes grayscale images. If a color image is processed, a robust reversible watermark can be embedded in one color channel, or the robust reversible watermark can be repeatedly embedded in three color channels. If a robust reversible watermark is embedded in one color channel, it cannot be guaranteed that the color image is completely protected; if a robust reversible watermark is repeatedly embedded in three color channels, the algorithm has low time efficiency and has certain robustness deficiencies. Therefore, how to embed a robust reversible watermark in a color image is a technical problem that needs to be solved urgently by those skilled in the art. Summary of the invention

[0004] In order to solve the above technical problems, the present invention proposes a color image robust reversible watermarking method based on quaternion Zernike moments, which represents and processes the pixels of the color image as a whole and has strong robustness against noise attacks, filter attacks, and JPEG compression.

[0005] The technical solution adopted by the present invention is: a color image robust reversible watermarking method based on quaternion Zernike moments, comprising the following steps:

[0006] Step 1: Construction of watermark embedding domain: The watermark embedding domain is constructed by normalizing the embedding coefficients when embedding the watermark in the quaternion Zernike moment of the original color image;

[0007] Step 2: Watermark embedding: embedding a robust watermark on the original color image;

[0008] Step 3: Generate an intermediate image: Generate an intermediate image close to the original color image based on the image embedded with the robust watermark;

[0009] Step 4: Calculate the error and moment coefficient error between the intermediate image and the original color image, and use the prediction error expansion reversible information hiding method to embed the above two errors into the image with the robust watermark added, and generate the final image embedded with the robust reversible watermark;

[0010] Step 5: watermark extraction;

[0011] Step 6: Restoration of the original color image.

[0012] The implementation process of step 1 is as follows:

[0013] 1a): Calculate the quaternion Zernike moment of the original color image;

[0014] 1b): Select moment coefficients and perform standardization: Select quaternion Zernike moment coefficients according to the capacity of the watermark, and perform standardization on the selected quaternion Zernike moment coefficients.

[0015] The implementation process of step 2 is as follows:

[0016] 2a): Scrambling encryption of watermark: Select a meaningful binary image as the watermark, and perform Arnold scrambling on the watermark to eliminate the spatial relationship of the watermark;

[0017] 2b): Robustly embed the watermark into the selected standardized quaternion Zernike moment coefficients;

[0018] 2c): Construct the perturbation f between the original color image and the image embedded with the robust watermark error ;

[0019] 2d): Get the image f with watermark w added w is: f w =uint8(f+f error ), where f is the original color image.

[0020] The implementation process of step 4 is as follows:

[0021] 4a): The intermediate image f obtained in step 3 w_remove , calculate the error f between the intermediate image and the original color image w_remove_error , where f w_remove_error =ff w_remove ;

[0022] 4b): Calculate the watermark image f w The moment coefficient error Z error ;

[0023] 4c): f w_remove_error and Z error Convert to binary sequence, denoted as Idate ;

[0024] 4d):I date The reversible information hiding method based on prediction error expansion is used to embed the watermark image f w The final color image f embedded with a robust reversible watermark is obtained w_final .

[0025] The implementation process of step 5 is as follows:

[0026] 5a): Calculate f w_final The quaternion Zernike moment of And perform the same standardization process as step 1b) to obtain the standardized quaternion Zernike moment

[0027] 5b): From Extract the watermark, assuming that the extracted watermark is N is the total number of bits of the watermark;

[0028] 5c): Yes Implement inverse Arnold transform to get a meaningful watermark.

[0029] The implementation process of step 6 is as follows:

[0030] 6a): From f w_final Extract I date Binary sequence, get remove I date The watermark image f w ;

[0031] 6b): I date Convert the binary sequence to a real number f w_remove_error and Z error ;

[0032] 6c): According to Z error Calculate f error ;

[0033] 6d): According to f error and f w , generate fw_remove;

[0034] 6e): According to fw_remove and f w_remove_error Get the original color image.

[0035] The expression of fw_remove generated in step 6d) is: w_remove =uint8(f w -f error ).

[0036] The expression of the original color image obtained in step 6e) is: f = f w_remove_error +f w_remove .

[0037] Compared with the prior art, the present invention has the following beneficial effects: Based on the quaternion representation method of color images, the three imaginary parts of the quaternion represent the red, green and blue color channels of the color image respectively, and the robust reversible watermark is embedded in the quaternion transformation domain of the color image. When extracting the watermark, it has strong robustness against filtering attacks, noise and JPEG compression; and, while ensuring that the original color image can be restored, the robustness of the embedded watermark is further improved, which can be applied to the copyright protection of color medical images. Specifically:

[0038] (1) Quaternions are used to represent color images, and robust reversible watermarks are embedded in the quaternion Zernike moments of color images, ensuring the overall processing of color image pixels in the robust reversible watermarking algorithm.

[0039] (2) The watermark is embedded and reversible information is hidden by combining quaternion moment coefficient quantization modulation, which realizes the robust reversible watermark embedding into color images. It is more efficient than processing each color channel of the color image separately.

[0040] (3) A robust reversible watermark is embedded in the quaternion Zernike moment coefficients of the color image to ensure the robustness of the algorithm.

[0041] (4) It can be applied to copyright protection of color medical images. After extracting the watermark and confirming the copyright, the original color image can be restored from the image embedded with the robust reversible watermark, and the accuracy of medical image diagnosis is guaranteed.

[0042] (5) In the cloud storage scenario of color medical images, the source information of the image is embedded into the image data as a watermark. The source of the image can be identified during the image transmission process, which can not only enhance the security of the medical image data, but also restore the original medical image for diagnosis when it is displayed. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] The present invention will be further described below in conjunction with the accompanying drawings:

[0044] Figure 1 The embedding process of the robust reversible watermark according to the embodiment of the present invention;

[0045] Figure 2 The watermark extraction and image restoration process of the embodiment of the present invention;

[0046] Figure 3 An example of embedding a robust reversible watermark in an image according to an embodiment of the present invention;

[0047] Figure 4 This is an example of extracting a watermark and restoring the original color image from an image embedded with a robust reversible watermark according to an embodiment of the present invention. DETAILED DESCRIPTION

[0048] like Figures 1 to 4 As shown, the present invention provides a color image robust reversible watermarking method based on quaternion Zernike moment, comprising the following steps:

[0049] Step 1: Construction of watermark embedding domain. This step realizes the standardization of embedding coefficients when embedding watermarks in the quaternion Zernike moments of color images. The specific implementation steps are as follows:

[0050] 1a): Calculate the quaternion Zernike moment of the original color image;

[0051] Let f represent the original color image, with a size of M×M. The quaternion representation of f is:

[0052] f=f r i+f g j+f b k;

[0053] Where: f r ,f g ,f b Respectively represent the red, green and blue color channels of the original color image, and i, j and k are the three imaginary units of the quaternion. Then the quaternion Zernike moment of the original color image is expressed as:

[0054]

[0055] Among them, r is the radius of the image in polar coordinates, θ is the phase angle, is a pure unit quaternion, R nm (r) is a Zernike polynomial, expressed as:

[0056]

[0057] Among them: the integer n represents the order of the Zernike moment, the integer m represents the repeatability of the Zernike moment, and it must satisfy |m|≤n|.

[0058] 1b): Select moment coefficients and perform standardization. Set an integer T according to the capacity of the watermark, select the quaternion Zernike moment coefficients of n+|m|≤T, and perform the following standardization on the selected moment coefficients:

[0059]

[0060] in, is the amplitude of the standardized quaternion Zernike moment, c1=c start -γ×(|n|+|m|), where c start ,γ is a parameter,, has no physical meaning, and the control parameter c1 is set according to the experiment. |Z 0,0 | is the magnitude of the quaternion Zernike moment when the order is 0 and the repeatability is 0. |Z n,m | is the magnitude of the quaternion, which is related to Z n,m There are the following relationships:

[0061]

[0062] Among them, R is the real part of the quaternion, V is the three imaginary parts of the quaternion, and φ is the phase angle, which is expressed as:

[0063] φ=arctan(V / R).

[0064] Step 2: Embedding of watermark. The watermark is robustly embedded into the quaternion Zernike moment coefficients of the color image to obtain the watermark image f w The specific implementation steps are as follows:

[0065] 2a): Scrambling encryption of watermark. A meaningful binary image is selected as the watermark. In order to eliminate the spatial relationship of the watermark, the watermark is scrambled by Arnold to obtain the watermark w i ,i=1,2,...,N (N is the total number of bits of the watermark).

[0066] 2b): The watermark is robustly embedded into the selected normalized moment coefficients using the following embedding method:

[0067]

[0068] in, is the amplitude of the normalized quaternion Zernike moment after embedding the watermark, Δ is the quantization step size, which is set according to the experiment; floor(·) is the rounding down operation, d n,m is the rounding error of the amplitude of the normalized quaternion Zernike moment,

[0069] 2c): Construct the perturbation f of the original color image and the watermarked image error , which is expressed as follows:

[0070]

[0071] Where: M is the highest order of the quaternion Zernike moment selected for watermark embedding, m is the repeatability of the quaternion Zernike moment, which must satisfy |m|≤n|, and V is the quaternion zn,m The three imaginary parts of φ are the quaternion z n,m Phase angle.

[0072] 2d): Get the image f with watermark w added w for:

[0073] f w =uint8(f+f error ).

[0074] Step 3: Generate intermediate image f w_remove In order to restore the original color image when extracting the watermark, according to the watermark image f w Generate an intermediate image f that is close to the original color image w_remove .in

[0075] Step 4: Calculate the error between the intermediate image and the original color image, the moment coefficient error, and use the reversible information hiding method of prediction error expansion to embed f w , generate the final image embedded with a robust reversible watermark. The specific implementation steps are as follows:

[0076] 4a): Based on the intermediate image obtained in step 3, calculate the error f between the intermediate image and the original color image w_remove_error .

[0077] f w_remove_error =ff w_remove .

[0078] 4b): Calculate the watermark image f w The moment coefficient error Z error , which is expressed as follows:

[0079]

[0080] 4c): f w_remove_error and Z error Convert to binary sequence, denoted as I date .

[0081] 4d):I date The reversible information hiding method based on prediction error expansion is used to embed the watermark image f w , and finally obtain the color image f embedded with robust reversible watermark w_final .

[0082] Step 5: Extraction of watermark. The watermark can be extracted from the image that is finally embedded with the robust reversible watermark to identify the copyright of the image. The specific implementation steps are as follows:

[0083] 5a): Calculate f w_finalThe quaternion Zernike moment of And perform the same standardization process as step 1b) to obtain the standardized quaternion Zernike moment

[0084] 5b): From Extract the watermark, assuming that the extracted watermark is (N is the total number of bits of the watermark), then:

[0085]

[0086] Among them, D1 and D2 are The amplitude The errors generated by the quantization operation when embedding 0-bit and 1-bit information, D1 and D2 are calculated as follows:

[0087]

[0088] 5c): Yes Implement inverse Arnold transform to get a meaningful watermark.

[0089] Step 6: Restoration of the original color image. The error of the intermediate image and the moment coefficient error can be restored from the image finally embedded with the robust reversible watermark, and then the intermediate image can be derived based on the moment coefficient error. The original color image can be restored based on the intermediate image and the error of the intermediate image. The specific implementation steps are as follows:

[0090] 6a): From f w_final Extract I date Binary sequence, get remove I date The watermark image f w .

[0091] 6b): I date Convert the binary sequence to a real number f w_remove_error and Z error .

[0092] 6c): According to Z error Calculate f error , and its calculation formula is as follows:

[0093]

[0094] 6d): According to f error and f w , generating f w_remove , and its calculation formula is as follows:

[0095] f w_remove =uint8(f w -f error );

[0096] 6e): Generate the original color image f. The original color image can be obtained according to f w_remove and f w_remove_error The calculation method is:

[0097] f=f w_remove_error +f w_remove .

[0098] The present invention will be further described below based on the specific embodiments shown in the accompanying drawings.

[0099] The colored pepper image is taken as the original color image, whose size is 256×256, and the binary image fawn image is taken as the watermark, whose size is 32×32. The parameters in the algorithm are set as: M=73, Δ=50, c start =5000,γ=30.

[0100] like Figure 3 As shown in the figure, (a) is the original color image, (b) is the watermark, (c) is the perturbation amount of the image (the pixel value is magnified 30 times to display), and (d) is the image after embedding the robust reversible watermark.

[0101] Image with robust reversible watermark (see attached) Figure 3 (d) is shown in Figure 1 The robust reversible watermark embedding process is obtained. The added watermark is a binary image of a deer (as shown in the attached Figure 3 (b)).

[0102] like Figure 4 FIG. 1 is an example of extracting a watermark and restoring an original color image from an image embedded with a robust reversible watermark according to an embodiment of the present invention.

[0103] According to the watermark extraction and image recovery process (such as Figure 2 As shown), it is possible to extract the watermark and restore the original color image from the image embedded with the robust reversible watermark (as shown Figure 4 As shown), when the image embedded with the robust reversible watermark is not attacked, the watermark can be extracted (as shown Figure 4 (b)) and restore the original color image (e.g. Figure 4 (c)).

[0104] Salt and pepper noise, Gaussian noise, JPEG image compression, image rotation, image scaling, and Gaussian blur are added to the pepper image embedded with a robust reversible watermark, and then the watermark is extracted using the degraded image. The watermark information can be clearly identified from the extracted watermark (the results are shown in Table 1).

[0105]

[0106] Table 1.

[0107] It can be seen that the present invention has strong robustness against these image attack operations.

[0108] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A robust reversible watermarking method for color images based on quaternion Zernike moments, characterized by: The following steps are involved: Step 1: Construction of watermark embedding domain: The watermark embedding domain is constructed by normalizing the embedding coefficients when embedding the watermark in the quaternion Zernike moment of the original color image; Step 2: Watermark embedding: embedding a robust watermark on the original color image; Step 3: Generate an intermediate image: Generate an intermediate image close to the original color image based on the image embedded with the robust watermark; Step 4: Calculate the error and moment coefficient error between the intermediate image and the original color image, and use the prediction error expansion reversible information hiding method to embed the above two errors into the image with the robust watermark added, and generate the final image embedded with the robust reversible watermark; Step 5: watermark extraction; Step 6: Restoration of the original color image.

2. According to claim 1, a color image robust reversible watermarking method based on quaternion Zernike moments is characterized by: The implementation process of step 1 is as follows: 1a): Calculate the quaternion Zernike moment of the original color image; 1b): Select moment coefficients and perform standardization: Select quaternion Zernike moment coefficients according to the capacity of the watermark, and perform standardization on the selected quaternion Zernike moment coefficients.

3. According to claim 2, a color image robust reversible watermarking method based on quaternion Zernike moments is characterized by: The implementation process of step 2 is as follows: 2a): Scrambling encryption of watermark: Select a meaningful binary image as the watermark, and perform Arnold scrambling on the watermark to eliminate the spatial relationship of the watermark; 2b): Robustly embed the watermark into the selected standardized quaternion Zernike moment coefficients; 2c): Construct the perturbation f between the original color image and the image embedded with the robust watermark error ; 2d): Get the image f with watermark w added w is: f w =uint8(f+f error ), where f is the original color image.

4. The method for robust reversible watermarking of color images based on quaternion Zernike moments according to claim 3 is characterized in that: The implementation process of step 4 is as follows: 4a): The intermediate image f obtained in step 3 w_remove , calculate the error f between the intermediate image and the original color image w_remove_error , where f w_remove_error =ff w_remove ; 4b): Calculate the watermark image f w The moment coefficient error Z error ; 4c): f w_remove_error and Z error Convert to binary sequence, denoted as I date ; 4d):I date The reversible information hiding method based on prediction error expansion is used to embed the watermark image f w The final color image f embedded with a robust reversible watermark is obtained w_final .

5. The method for robust reversible watermarking of color images based on quaternion Zernike moments according to claim 4 is characterized in that: The implementation process of step 5 is as follows: 5a): Calculate f w_final The quaternion Zernike moment of And perform the same standardization process as step 1b) to obtain the standardized quaternion Zernike moment 5b): From Extract the watermark, assuming that the extracted watermark is i=1,2,...,N, N is the total number of bits of the watermark; 5c): Yes Implement inverse Arnold transform to get a meaningful watermark.

6. The method for robust reversible watermarking of color images based on quaternion Zernike moments according to claim 5, characterized in that: The implementation process of step 6 is as follows: 6a): From f w_final Extract I date Binary sequence, get remove I date The watermark image f w ; 6b): I date Convert binary sequence to real number f w_remove_error and Z error ; 6c): According to Z error Calculate f error ; 6d): According to f error and f w , generating f w_remove ; 6e): According to f w_remove and f w_remove_error Get the original color image.

7. The method for robust reversible watermarking of color images based on quaternion Zernike moments according to claim 6, characterized in that: f generated in step 6d) w_remove The expression of f is: w_remove =uint8(f w -f error ).

8. The robust reversible watermarking method for color images based on quaternion Zernike moments according to claim 6 is characterized by: The expression of the original color image obtained in step 6e) is: f = f w_remove_error +f w_remove .

Images