Reversible information hiding prediction method and device and storage medium

Abstract

The invention provides a reversible information hiding prediction method and device and a storage medium, and relates to the technical field of reversible information hiding. According to the method, an original image is divided into a point set image and a star set image by setting a segmentation mode, an initial predictor is constructed and trained to obtain a prediction optimizer, the predicted star set image and the predicted point set image are combined to obtain a predicted image, floating value calculation is carried out on pixels of the predicted image, and texture complexity is calculated according to the floating value calculation. According to the method, the texture complexity of the image is calculated, the pixels are sorted according to the texture complexity to obtain the secret-carrying image, the texture features of the image can be deeply analyzed, a quantitative basis is provided for subsequent processing, reasonable processing of the image pixels is achieved, and damage to the visual quality of the image is effectively reduced while information hiding is guaranteed.

Description

Technical Field

[0001] The present invention mainly relates to the technical field of reversible information hiding, and in particular to a reversible information hiding prediction method, device and storage medium. Background Art

[0002] With the increasing popularity of the Internet, people's ability to obtain, disseminate and process information has become more and more convenient. At the same time, some people use the Internet to maliciously steal and tamper with digital media information, which not only dampens the enthusiasm of creators, but also infringes on copyright. In order to protect the copyright of digital media, reversible information hiding technology has been proposed. Reversible information hiding technology not only has high imaging performance, but also can restore the original image information losslessly, ensuring that the original image is not destroyed.

[0003] In the existing reversible information hiding technology, the image needs to be preprocessed before embedding information, and then the information is embedded in the image. By processing the image with a prediction algorithm, after embedding the data, the embedding performance of the image can be improved, the distortion rate of the image can be reduced, and the imaging quality can be improved. However, the existing technology does not achieve a good balance between improving the embedding capacity and reducing the distortion rate of the image, which is easy to cause image distortion. Summary of the invention

[0004] The technical problem to be solved by the present invention is to provide a reversible information hiding prediction method, device and storage medium in view of the deficiencies in the prior art.

[0005] The technical solution of the present invention to solve the above technical problem is as follows: a reversible information hiding prediction method, comprising the following steps:

[0006] S1. Segment the original image according to the set segmentation method to obtain a point set image and a star set image;

[0007] S2, construct an initial predictor, and train the initial predictor through the initial training set to obtain a prediction optimizer;

[0008] S3, training the point set image through the prediction optimizer to output a predicted star set image, and training the star set image through the prediction optimizer to output a predicted point set image;

[0009] S4, combining the predicted star set image and the predicted point set image to obtain a predicted image;

[0010] S5, performing floating value calculation on pixels of the predicted image, and calculating the texture complexity of the predicted image according to the floating value;

[0011] S6. Sort the pixels of the predicted image according to the texture complexity, and obtain a dense image through a pixel sequence obtained by sorting.

[0012] Another technical solution of the present invention to solve the above technical problem is as follows: a reversible information hiding prediction device, comprising:

[0013] A segmentation module is used to segment the original image according to a set segmentation method to obtain a point set image and a star set image;

[0014] A predictor construction module is used to construct an initial predictor and train the initial predictor through an initial training set to obtain a prediction optimizer;

[0015] A training module, used to train the point set image through the prediction optimizer and output a predicted star set image, and to train the star set image through the prediction optimizer and output a predicted point set image;

[0016] A combination module, used for combining the predicted star set image and the predicted point set image to obtain a predicted image;

[0017] A calculation and analysis module, used for performing floating value calculation on pixels of the predicted image, and calculating the texture complexity of the predicted image according to the floating value;

[0018] The dense image acquisition module is used to sort the pixels of the predicted image according to the texture complexity, and obtain the dense image through the pixel sequence obtained by sorting.

[0019] Another technical solution of the present invention to solve the above technical problem is as follows: a reversible information hiding prediction device includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the reversible information hiding prediction method as described above is implemented.

[0020] Another technical solution of the present invention to solve the above technical problem is as follows: a reversible information hiding prediction device includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the reversible information hiding prediction method as described above is implemented.

[0021] The beneficial effects of the present invention are as follows: by setting a segmentation method to divide the original image into a point set image and a star set image, it is helpful to carry out targeted processing of image contents with different features, improve the accuracy and efficiency of subsequent processing, construct and train an initial predictor to obtain a prediction optimizer, and continuously optimize the prediction capability, so that the prediction of the image is more accurate, laying a foundation for the subsequent generation of high-quality predicted images, combining the predicted star set image and the predicted point set image to obtain a predicted image, making full use of the features of different set images, so that the predicted image more completely and accurately reflects the information of the original image, performing floating value calculation on the pixels of the predicted image and calculating the texture complexity accordingly, and then sorting the pixels according to the texture complexity to obtain a encrypted image, which can deeply analyze the texture features of the image, provide a quantitative basis for subsequent processing, and realize reasonable processing of image pixels, while ensuring information hiding, effectively reducing the damage to the visual quality of the image. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 A flowchart of a reversible information hiding prediction method provided by an embodiment of the present invention;

[0023] Figure 2 A module block diagram of a reversible information hiding prediction device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0024] The principles and features of the present invention are described below in conjunction with the accompanying drawings. The examples given are only used to explain the present invention and are not used to limit the scope of the present invention.

[0025] like Figure 1 As shown, an embodiment of the present invention provides a reversible information hiding prediction method, comprising the following steps:

[0026] S1. Segment the original image according to the set segmentation method to obtain a point set image and a star set image;

[0027] S2, construct an initial predictor, and train the initial predictor through the initial training set to obtain a prediction optimizer;

[0028] S3, training the point set image through the prediction optimizer to output a predicted star set image, and training the star set image through the prediction optimizer to output a predicted point set image;

[0029] S4, combining the predicted star set image and the predicted point set image to obtain a predicted image;

[0030] S5, performing floating value calculation on pixels of the predicted image, and calculating the texture complexity of the predicted image according to the floating value;

[0031] S6. Sort the pixels of the predicted image according to the texture complexity to obtain a pixel sequence {bc1, bc2, ..., bc i ,…}, and obtain the encrypted image by sorting the pixel sequence.

[0032] It should be understood that the initial predictor is constructed based on the CNN predictor, and an optimization algorithm is added to the CNN predictor to obtain the prediction optimizer.

[0033] The CNN prediction optimizer is trained using the initial predictor training set. This optimized predictor is trained on the point set image and outputs the predicted star set image. Similarly, the star set is trained to obtain the prediction of the point set.

[0034] In the above embodiment, the original image is divided into a point set image and a star set image by setting a segmentation method, which helps to perform targeted processing on image content with different features, improve the accuracy and efficiency of subsequent processing, build and train the initial predictor to obtain a prediction optimizer, which can continuously optimize the prediction ability, make the prediction of the image more accurate, and lay the foundation for the subsequent generation of high-quality predicted images. The predicted star set image and the predicted point set image are combined to obtain a predicted image, which makes full use of the features of different set images, so that the predicted image more completely and accurately reflects the information of the original image. The pixels of the predicted image are calculated by floating values ​​and the texture complexity is calculated accordingly. Then, the pixels are sorted according to the texture complexity to obtain a encrypted image, which can deeply analyze the texture features of the image, provide a quantitative basis for subsequent processing, and realize reasonable processing of image pixels. While ensuring information hiding, it effectively reduces the damage to the visual quality of the image.

[0035] Preferably, the step S1 is specifically:

[0036] The original image is segmented in a chessboard segmentation manner to obtain a point set image and a star set image.

[0037] In the above embodiment, the point set and the star set may represent different feature areas in the image respectively, which can make prediction and information embedding more accurate and improve the accuracy of the overall processing.

[0038] Preferably, in step S3, the point set image is trained by the prediction optimizer to output a predicted star set image, specifically:

[0039] Point set image I a Perform feature extraction, that is, perform convolution operation to obtain image I b1 ;

[0040] For the image I b1 Perform global prediction and obtain the predicted image Ib2 ;

[0041] The predicted image I is optimized by b2 Optimize and get the optimized image I′ b , the optimization formula is:

[0042] Θ1=(p1+p2+p3+p4) / 4,

[0043] Among them, Θ1 is the optimized image I′ b In the optimized target pixel, p1, p2, p3 and p4 are the four diagonal pixels in the target pixel respectively;

[0044] Training loss function, the loss function is:

[0045]

[0046] Among them, I′ b To predict the star set image, I a is the point set image used for prediction, μ is the weight of the convolution kernel, and ξ is the bias;

[0047] After the training is completed, the prediction optimizer is obtained and the predicted star set image is output.

[0048] Preferably, in step S3, the star set image is trained by the prediction optimizer to output a predicted point set image, specifically:

[0049] Star collection image I b Perform feature extraction, that is, perform convolution operation to obtain image I a1 ;

[0050] For image I a1 Perform global prediction and obtain the predicted image I a2 ;

[0051] The predicted image I is optimized by a2 Optimize and get the optimized image I′ a , the optimization formula is:

[0052] Θ2=(p1+p2+p3+p4) / 4,

[0053] Among them, Θ2 is the optimized image I′ a In the optimized target pixel, p1, p2, p3 and p4 are the four diagonal pixels in the target pixel respectively;

[0054] Training loss function, the loss function is:

[0055]

[0056] Among them, I′ a To predict the posterior point set image, I b is the star set image used for prediction, μ is the weight of the convolution kernel, and ξ is the bias;

[0057] After training is completed, the predicted point set image is output.

[0058] In the above embodiment, the initial predictor is constructed and trained to obtain a prediction optimizer, which can continuously optimize the prediction capability, making the prediction of the image more accurate, and laying a foundation for the subsequent generation of high-quality predicted images.

[0059] The output predicted star set image and predicted point set image are combined to obtain a predicted image, which is close to the original image to a large extent, provides a better basis for subsequent information embedding, and reduces the distortion of the image caused by information embedding.

[0060] It should be understood that “combining” means stitching together the two images, the predicted star set image and the predicted point set image.

[0061] Preferably, in step S5, floating value calculation is performed on pixels of the predicted image, specifically:

[0062] The floating values ​​of the predicted star set image and the predicted point set image in the predicted image are respectively calculated by a floating value calculation formula, and the floating value calculation formula is:

[0063] Δ P =|ab|+|bc|+|cd|+|da|+|ac|+|bd|,

[0064] Among them, Δ P is the floating value corresponding to the predicted star set image or the predicted point set image, a, b, c and d are the upper, left, lower and right contexts corresponding to any pixel in the predicted star set image or the predicted point set image, that is, the horizontal, vertical and adjacent pixel differences of the four pixels a, b, c and d.

[0065] Preferably, in step S5, floating value calculation is performed on pixels of the predicted image, specifically:

[0066] In step S5, the texture complexity of the predicted image is calculated according to the floating value, specifically:

[0067] The texture complexity of the predicted star set image and the predicted point set image are calculated respectively by using the texture complexity formula and the floating value, and the texture complexity formula is:

[0068]

[0069] in, is the texture complexity corresponding to the predicted star set image or the predicted point set image, n is the number of available texture complexities around any texture complexity, and n∈[1,4].

[0070] In the above embodiment, by calculating the floating value and texture complexity of the predicted image pixels, the beneficial effect is that the texture complexity can be further calculated according to the characteristics of the pixels (such as the context pixel differences reflected by the floating value), thereby providing a more accurate basis for the subsequent sorting of pixels according to the texture complexity to obtain the secret image, which helps to better process the image pixels while ensuring the information hiding, reduce the damage to the image visual quality, and realize the effective processing of the image and information hiding.

[0071] This method provides a systematic reversible information hiding scheme. From image segmentation, predictor training, error analysis to information embedding, each step cooperates with each other and can effectively realize information hiding and reversible restoration of the original image.

[0072] For application scenarios such as digital media copyright protection, this method can embed important data such as copyright information into images without significantly affecting image quality, thereby improving the security and traceability of digital media.

[0073] like Figure 2 As shown, the present invention also provides a reversible information hiding prediction device, comprising:

[0074] A segmentation module is used to segment the original image according to a set segmentation method to obtain a point set image and a star set image;

[0075] A predictor construction module is used to construct an initial predictor and train the initial predictor through an initial training set to obtain a prediction optimizer;

[0076] A training module, used to train the point set image through the prediction optimizer and output a predicted star set image, and to train the star set image through the prediction optimizer and output a predicted point set image;

[0077] A combination module, used for combining the predicted star set image and the predicted point set image to obtain a predicted image;

[0078] A calculation and analysis module, used for performing floating value calculation on pixels of the predicted image, and calculating the texture complexity of the predicted image according to the floating value;

[0079] The dense image acquisition module is used to sort the pixels of the predicted image according to the texture complexity, and obtain the dense image through the pixel sequence obtained by sorting.

[0080] The computing and analysis module is specifically used for:

[0081] The floating values ​​of the predicted star set image and the predicted point set image in the predicted image are respectively calculated by a floating value calculation formula, and the floating value calculation formula is:

[0082] Δ P =|ab|+|bc|+|cd|+|da|+|ac|+|bd|,

[0083] Among them, Δ P is the floating value corresponding to the predicted star set image or the predicted point set image, a, b, c and d are the upper, left, lower and right contexts corresponding to any pixel in the predicted star set image or the predicted point set image, that is, the horizontal, vertical and adjacent pixel differences of the four pixels a, b, c and d.

[0084] The present invention also provides a reversible information hiding prediction device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the reversible information hiding prediction method described above is implemented.

[0085] The present invention also provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the reversible information hiding prediction method as described above is implemented.

[0086] The present invention has significant advantages in the field of reversible information hiding technology through a specific reversible information hiding prediction method, device and storage medium. The specific beneficial effects are as follows:

[0087] Improve prediction accuracy: The prediction optimizer is used to train the point set image and the star set image respectively, and the predicted point set image and the predicted point set image are output. Finally, the two are combined to obtain the predicted image, so that the image content can be predicted more accurately.

[0088] Balance information hiding and visual quality: While ensuring information hiding, it effectively reduces the damage to the visual quality of the image.

[0089] Optimize image pixel processing: By performing floating value calculations on predicted image pixels, calculating texture complexity based on the floating values, and then sorting pixels according to texture complexity to obtain a dense image, reasonable processing of image pixels is achieved.

[0090] Adaptability to complex original images: The entire method can effectively process complex original images, which improves the applicability of the method.

[0091] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.

[0092] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described devices and units can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0093] In the several embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic, for example, the division of units is only a logical function division, and there may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed.

[0094] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the embodiments of the present invention.

[0095] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0096] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods of each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk and other media that can store program codes.

[0097] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A reversible information hiding prediction method, characterized in that: The steps include: S1. Segment the original image according to the set segmentation method to obtain a point set image and a star set image; S2, construct an initial predictor, and train the initial predictor through the initial training set to obtain a prediction optimizer; S3, training the point set image through the prediction optimizer to output a predicted star set image, and training the star set image through the prediction optimizer to output a predicted point set image; S4, combining the predicted star set image and the predicted point set image to obtain a predicted image; S5, performing floating value calculation on pixels of the predicted image, and calculating the texture complexity of the predicted image according to the floating value; S6. Sort the pixels of the predicted image according to the texture complexity, and obtain a dense image through a pixel sequence obtained by sorting.

2. The prediction method according to claim 1, characterized in that: In step S3, the point set image is trained by the prediction optimizer to output a predicted star set image, specifically: Point set image I a Perform feature extraction to obtain image I b1 ; For the image I b1 Perform global prediction and obtain the predicted image I b2 ; The predicted image I is optimized by b2 Optimize and get the optimized image I′ b , the optimization formula is: Θ1=(p1+p2+p3+p4) / 4, Among them, Θ1 is the optimized image I′ b In the optimized target pixel, p1, p2, p3 and p4 are the four diagonal pixels in the target pixel respectively; Training loss function, the loss function is: Among them, I′ b To predict the star set image, I a is the point set image used for prediction, μ is the weight of the convolution kernel, and ξ is the bias; After the training is completed, the prediction optimizer is obtained and the predicted star set image is output.

3. The prediction method according to claim 1, characterized in that: In step S3, the prediction optimizer is used to train the star set image and output a predicted point set image, specifically: Star collection image I b Perform feature extraction to obtain image I a1 ; For image I a1 Perform global prediction and obtain the predicted image I a2 ; The predicted image I is optimized by a2 Optimize and get the optimized image I′ a , the optimization formula is: Θ2=(p1+p2+p3+p4) / 4, Among them, Θ2 is the optimized image I′ a In the optimized target pixel, p1, p2, p3 and p4 are the four diagonal pixels in the target pixel respectively; Training loss function, the loss function is: Among them, I′ a To predict the posterior point set image, I b is the star set image used for prediction, μ is the weight of the convolution kernel, and ξ is the bias; After training is completed, the predicted point set image is output.

4. The prediction method according to claim 1, characterized in that: In step S5, floating value calculation is performed on the pixels of the predicted image, specifically: The floating values ​​of the predicted star set image and the predicted point set image in the predicted image are respectively calculated by a floating value calculation formula, and the floating value calculation formula is: Δ P =|a-b|+|b-c|+|c-d|+|d-a|+|a-c|+|b-d|, Among them, Δ P is the floating value corresponding to the predicted star set image or the predicted point set image, a, b, c and d are the upper, left, lower and right contexts corresponding to any pixel in the predicted star set image or the predicted point set image, that is, the horizontal, vertical and adjacent pixel differences of the four pixels a, b, c and d.

5. The prediction method according to claim 4, characterized in that: In step S5, the texture complexity of the predicted image is calculated according to the floating value, specifically: The texture complexity of the predicted star set image and the predicted point set image are calculated respectively by using the texture complexity formula and the floating value, and the texture complexity formula is: in, is the texture complexity corresponding to the predicted star set image or the predicted point set image, n is the number of available texture complexities around any texture complexity, and n∈[1,4].

6. The prediction method according to any one of claims 1 to 5, characterized in that: The step S1 is specifically: The original image is segmented in a chessboard segmentation manner to obtain a point set image and a star set image.

7. A reversible information hiding prediction device, characterized in that: include: A segmentation module is used to segment the original image according to a set segmentation method to obtain a point set image and a star set image; A predictor building module is used to build an initial predictor and train the initial predictor through an initial training set to obtain a prediction optimizer; A training module, used to train the point set image through the prediction optimizer and output a predicted star set image, and to train the star set image through the prediction optimizer and output a predicted point set image; A combination module, used for combining the predicted star set image and the predicted point set image to obtain a predicted image; A calculation and analysis module, used for performing floating value calculation on pixels of the predicted image, and calculating the texture complexity of the predicted image according to the floating value; The dense image acquisition module is used to sort the pixels of the predicted image according to the texture complexity, and obtain the dense image through the pixel sequence obtained by sorting.

8. The prediction device according to claim 7, characterized in that The computing and analysis module is specifically used for: The floating values ​​of the predicted star set image and the predicted point set image in the predicted image are respectively calculated by a floating value calculation formula, and the floating value calculation formula is: Δ P =|a-b|+|b-c|+|c-d|+|d-a|+|a-c|+|b-d|, Among them, Δ P is the floating value corresponding to the predicted star set image or the predicted point set image, a, b, c and d are the upper, left, lower and right contexts corresponding to any pixel in the predicted star set image or the predicted point set image, that is, the horizontal, vertical and adjacent pixel differences of the four pixels a, b, c and d.

9. A reversible information hiding prediction device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the reversible information hiding prediction method according to any one of claims 1 to 6 is implemented.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the reversible information hiding prediction method according to any one of claims 1 to 6 is implemented.

Images