An adaptive image watermark embedding and recovery system

Abstract

The application relates to the field of digital watermark technology and discloses a self-adaptive image watermark embedding and recovery system, which comprises an image preprocessing module, a watermark preprocessing module and a watermark embedding module.The image preprocessing module is used for carrying out discrete wavelet transform on a carrier image to be embedded with a watermark, decomposing the image into a plurality of frequency subbands, and selecting a low-frequency subband as a watermark embedding area.The watermark preprocessing module is used for carrying out orthogonal transformation on a watermark image, so that the plurality of watermarks keep orthogonality and mutual interference of the watermarks is reduced.The watermark embedding module is used for embedding the watermark into the low-frequency subband of the carrier image based on the output of the watermark preprocessing module.Through self-adaptive optimization of watermark strength, redundant embedding and orthogonal transformation technology, the application improves the concealment and robustness of the watermark.Compared with the prior art, the application effectively reduces the interference between the watermarks and enhances the resistance of the watermark to attacks such as compression, noise, cropping and rotation.

Description

Technical Field

[0001] This invention relates to the field of digital watermarking technology, specifically to an adaptive image watermark embedding and restoration system. Background Technology

[0002] In digital watermarking technology, watermark embedding and extraction are crucial means of protecting the copyright and integrity of digital content. Existing watermarking technologies are widely used to protect multimedia content such as images, videos, and audio, primarily achieving functions like copyright protection and information verification through embedded watermark information. However, despite advancements in existing technologies, many watermarking techniques still have significant shortcomings, particularly in areas such as the adaptability of watermark embedding strength, interference between multiple watermarks, robustness, and extraction accuracy.

[0003] In existing technologies, watermark embedding methods employ fixed watermark strength and position, neglecting the differences in image content. This fixed embedding method easily causes visual interference in complex and low-contrast areas of the image, increasing watermark visibility. Furthermore, many techniques fail to effectively address the mutual interference and overlap between watermarks when handling multiple watermarks, thus affecting extraction accuracy. While some techniques attempt to rely excessively on low-frequency components during embedding to ensure concealment, the watermarks exhibit poor robustness against attacks such as compression, noise interference, and cropping, making stable recovery impossible. More importantly, existing watermarking technologies lack sufficient control over extraction accuracy, often resulting in significant errors during recovery, especially after image compression or cropping, where the accuracy and stability of watermark extraction cannot be effectively guaranteed. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides an adaptive image watermark embedding and restoration system, which solves the problems of poor watermark concealment, weak anti-attack capability, and insufficient watermark restoration accuracy in existing technologies.

[0005] To achieve the above objectives, the present invention provides the following technical solution: an adaptive image watermark embedding and restoration system, comprising:

[0006] The image preprocessing module is used to perform discrete wavelet transform on the carrier image to be embedded with watermark, decompose the image into multiple frequency sub-bands, and select the low-frequency sub-band as the watermark embedding region.

[0007] The watermark preprocessing module is used to perform orthogonal transformation on the watermark image to maintain the orthogonality between multiple watermarks and reduce mutual interference between watermarks.

[0008] The watermark embedding module is used to embed the watermark into the low-frequency subband of the carrier image based on the output of the watermark preprocessing module, and adjust the embedding strength through singular value decomposition to match the watermark information with the features of the carrier image, thereby improving robustness and concealment.

[0009] The watermark optimization and enhancement module is used to optimize watermark embedding parameters. By dynamically adjusting the embedding strength, the watermark can maintain its concealment while improving information capacity and robustness, and reducing the similarity between watermarks.

[0010] The watermark extraction and restoration module is used to recover watermark information from images with embedded watermarks based on inverse discrete wavelet transform and inverse singular value decomposition. It also optimizes the extraction parameters based on optimal control theory, enabling it to adaptively optimize the extraction accuracy under different attack conditions and improve the stability of watermark restoration.

[0011] The robustness enhancement module is used to improve the watermark extraction success rate under JPEG compression, cropping, noise, and rotation attacks. It enhances the watermark's resistance to cropping through a spatially distributed embedding strategy, enhances the watermark's resistance to rotation through Fourier-Mellin transform, and enhances the watermark's resistance to noise through an adaptive filtering strategy.

[0012] The watermark optimization and enhancement module optimizes the watermark embedding strength based on the variational method and dynamically adjusts the embedding parameters through the optimization algorithm, so that the watermark can improve information capacity and robustness while ensuring concealment after embedding.

[0013] Preferably, the image preprocessing module performs discrete wavelet transform on the carrier image to generate low-frequency sub-bands and high-frequency sub-bands, and selects the low-frequency sub-bands as the watermark embedding region to improve the robustness of the watermark.

[0014] Preferably, the watermark preprocessing module uses orthogonal transformation to transform the watermark image, so that multiple watermarks maintain orthogonality and reduce mutual interference between watermarks, wherein the orthogonal transformation includes discrete cosine transform or discrete Fourier transform.

[0015] Preferably, the watermark embedding module decomposes the low-frequency sub-band of the carrier image through singular value decomposition and embeds the watermark information into the decomposed singular value matrix. The embedded singular value matrix is ​​used to reconstruct the low-frequency sub-band and is combined with inverse discrete wavelet transform to reconstruct the watermarked image.

[0016] Preferably, the watermark extraction and restoration module performs discrete wavelet transform on the watermarked image to extract the low-frequency sub-band, calculates the embedded watermark information based on singular value decomposition, and then restores the original watermark using inverse orthogonal transform.

[0017] Preferably, the watermark extraction and recovery module uses optimal control theory to dynamically adjust the watermark recovery parameters, enabling it to adaptively optimize extraction accuracy under different attack conditions and improve the stability of watermark recovery.

[0018] Preferably, the robustness enhancement module employs a quantization adaptive strategy to adjust the quantization step size of the watermark embedding in response to JPEG compression attacks, so that the watermark information can still be recovered after JPEG compression.

[0019] Preferably, the robustness enhancement module uses a spatially distributed embedding strategy to store watermark information in multiple regions, thereby enhancing the watermark's resistance to cropping attacks.

[0020] Preferably, the robustness enhancement module uses Fourier-Mellin transform to transform the embedded watermark information to enhance the robustness of the watermark against rotation attacks. In response to noise attacks, the robustness enhancement module adopts an adaptive filtering strategy, including Wiener filtering and median filtering, to reduce the impact of noise on watermark extraction, and combines redundant coding to improve the stability of watermark extraction.

[0021] This invention provides an adaptive image watermark embedding and restoration system. It has the following advantages:

[0022] 1. This invention employs adaptive watermark strength optimization technology, combining variational method and optimal control theory to dynamically adjust the watermark embedding strength and position, ensuring that while maximizing watermark capacity, the impact on image quality is minimized. Compared with the fixed watermark strength settings commonly used in existing technologies, the optimization scheme of this invention can automatically adjust watermark parameters according to image content, improving the concealment of the watermark and reducing visual interference.

[0023] 2. This invention preprocesses the watermark through orthogonal transformation, ensuring the orthogonality of multiple watermark information and avoiding mutual interference between watermarks. Compared with traditional watermarking technology, this invention can effectively improve the independence of the multi-watermark system, significantly improving the extraction accuracy of each watermark information. Especially in complex environments with multiple watermark embeddings, it exhibits good stability and reliability.

[0024] 3. This invention introduces redundant embedding technology, which distributes watermark information to multiple regions of the image. Even if the image has undergone operations such as cropping or rotation, the watermark can still be recovered from the remaining part. Compared with the existing technology where the watermark is embedded in a single region of the image, which is susceptible to cropping attacks, this invention significantly improves the watermark's resistance to cropping and robustness, ensuring the stability of the watermark in complex scenarios.

[0025] 4. This invention employs adaptive filtering and anti-compression enhancement techniques to effectively reduce the impact of noise and compression on watermark recovery. Existing technologies often suffer from low watermark extraction accuracy and are easily interfered with when faced with image compression or noise. By optimizing the embedding process and combining it with denoising processing, this invention can significantly improve the watermark recovery accuracy under compression and noise interference environments, ensuring efficient extraction of watermark information. Attached Figure Description

[0026] Figure 1 This is a system architecture diagram of the present invention. Detailed Implementation

[0027] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0028] Please see the appendix Figure 1 This invention provides an adaptive image watermark embedding and restoration system, comprising:

[0029] The image preprocessing module is used to perform discrete wavelet transform on the carrier image to be embedded with watermark, decompose the image into multiple frequency sub-bands, and select the low-frequency sub-band as the watermark embedding region.

[0030] In this embodiment, the image preprocessing module first uses Discrete Wavelet Transform (DWT) to decompose the input carrier image. DWT is a common signal processing method that can divide an image into different frequency components for better information processing. In this invention, the DWT transform decomposes the input carrier image... It is decomposed into four sub-bands, namely the low-frequency sub-band. Horizontal high-frequency subband Vertical high-frequency subband and diagonal high-frequency subband :

[0031]

[0032] in:

[0033] Subbands mainly contain overall structural information of the image, with lower visual contrast but concentrated energy;

[0034] and Subbands contain the edges, textures, and high-frequency details of an image, and are usually more susceptible to noise and compression.

[0035] In one possible implementation, DWT uses Haar wavelets as basis functions, which are relatively simple to compute and suitable for image processing. The Haar wavelet transform can be used to calculate each subband using the following formula:

[0036]

[0037] in, , These represent the row and column indices of the sub-band matrix, respectively. The pixel values ​​of the input image.

[0038] Generally, the watermark embedding region is selected in the low-frequency sub-band. This is because this part of the data has the least impact on the image and its energy is more concentrated, allowing the embedded watermark to resist compression and noise interference better. Meanwhile, the high-frequency sub-band... , , It is susceptible to filtering, quantization, and compression, and is not suitable as a watermark embedding area.

[0039] In some embodiments, to further reduce the impact of watermarks on image quality, the image preprocessing module also performs... Singular value decomposition (SVD) is performed on the sub-bands to extract key singular value information for watermark embedding. Specifically, for The subband undergoes SVD transformation, which can be represented as follows:

[0040]

[0041] in:

[0042] It is a left singular matrix containing orthogonal basis information;

[0043] This is a singular value matrix, containing the main energy information of the image;

[0044] It is a right singular matrix, and They have the same orthogonal properties.

[0045] As an alternative, only the singular value matrix is ​​extracted during the image preprocessing stage. Watermark embedding is performed while maintaining and The matrix remains unchanged, which ensures that the impact of the embedding process on the image structure is minimized.

[0046] In another possible implementation, to further improve the adaptability of the watermark embedding region, the image preprocessing module can also combine edge detection algorithms, such as the Canny operator or the Sobel operator, to pre-analyze the carrier image and determine the main structural regions of the image. Specifically, the Canny operator can calculate the gradient information of the image:

[0047]

[0048] in:

[0049] Represents the image at the pixel level Gradient strength at the location;

[0050] and These are the gradient values ​​in the horizontal and vertical directions, respectively, which can be calculated using a Sobel convolution kernel.

[0051] Before embedding the watermark, regions with lower gradients can be selected for watermark embedding to reduce the impact of the embedding on the image edge structure. This method can further optimize the watermark embedding position and improve the imperceptibility of the watermark.

[0052] In some embodiments, to improve computational efficiency, the image can be segmented. Specifically, the carrier image can be divided into multiple... The watermark is divided into small blocks, each of which undergoes DWT and SVD transformations separately. This method helps improve the local adaptability of the watermark, allowing the embedding strength of the watermark in different regions to be dynamically adjusted, thereby improving the overall performance of the system.

[0053] Overall, the core objective of the image preprocessing module is to extract the optimal watermark embedding region, ensuring the stability and concealment of the embedded watermark. Discrete wavelet transform decomposes the image into different frequency sub-bands, providing appropriate storage space for the watermark. Combining this with singular value decomposition further optimizes the embedding process, enabling the watermark information to be embedded with minimal distortion. For certain specialized applications, edge detection and adaptive block partitioning can be combined to further optimize the watermark embedding position, giving the entire system greater adaptability and stability.

[0054] The watermark preprocessing module is used to perform orthogonal transformation on the watermark image to maintain the orthogonality between multiple watermarks and reduce mutual interference between watermarks.

[0055] In general, the watermark preprocessing module first performs an orthogonal transformation on the watermark image. The main goal of the orthogonal transformation is to minimize the correlation between multiple watermarks, ensuring that each watermark is as independent as possible during embedding, avoiding interference and information overlap between watermarks. This has the advantage that when multiple watermarks are embedded in the same carrier image, they can effectively avoid mutual interference, thereby improving the watermark extraction accuracy. After orthogonal transformation, the watermark information will have higher anti-interference capabilities and better concealment.

[0056] In this embodiment, the watermark preprocessing module uses Discrete Cosine Transform (DCT) or Discrete Fourier Transform (DFT) as orthogonal transformation methods. Specifically, the watermark image... First, a DCT transform is performed to obtain the transform result in the frequency domain. It can be represented as:

[0057]

[0058] in, The original watermark image. The orthogonal transformation matrix is ​​used, with common transformations including DCT or DFT, depending on the implementation. Through orthogonal transformation, the watermark information is decomposed into different frequency components. Typically, low-frequency components contain basic image information, while high-frequency components contain more details. High-frequency watermark information is more stable; therefore, the watermark preprocessing module prioritizes processing low-frequency components to ensure the robustness of the watermark after embedding.

[0059] As an alternative, the watermark information can be further optimized after orthogonal transformation. Generally, by selecting an appropriate frequency range and the intensity of the transformed watermark, the mutual interference between the watermark and the image can be further reduced, improving the accuracy of watermark extraction. In this process, the high-frequency components of the watermark are suppressed, avoiding high-frequency noise interference with image quality and watermark extraction.

[0060] Specifically, the orthogonal transformation used in this embodiment includes the Discrete Cosine Transform (DCT), which is commonly used in image compression and image processing. DCT can transform image data from the spatial domain to the frequency domain and embed watermark information into the low-frequency components of the image, thereby enhancing its anti-interference capability. The watermark information after DCT transformation... As shown below:

[0061]

[0062] in, The pixel values ​​of the watermark image. and These are the height and width of the watermark image, respectively. and These are the transformed frequency coordinates. Through this transformation, the frequency components of the watermark image are decomposed into a set of coefficients. The low-frequency components retain the main information of the watermark, while the high-frequency components retain more details.

[0063] In another possible implementation, to further improve the independence of the watermark, the watermark preprocessing module can also incorporate **Singular Value Decomposition (SVD)** to process the watermark information. SVD is a matrix factorization method that decomposes the watermark information matrix into three matrices: UUU, SSS, and VVV, where SSS is the matrix containing singular values, representing the energy and information of the watermark image. By performing SVD processing on the watermark, we can obtain the main information of the watermark image, thereby optimizing the subsequent embedding process.

[0064] Specifically, watermarked images It can be decomposed into:

[0065]

[0066] in:

[0067] and It is an orthogonal matrix, representing the feature vector of the watermark image;

[0068] It is a diagonal matrix containing the singular values ​​of the watermarked image.

[0069] During processing, the watermark preprocessing module adjusts and optimizes the singular value matrix (SSS) of the watermark to further enhance the watermark image's anti-interference capability and extraction accuracy. Watermark images processed in this way are more suitable for embedding because they are more independent and less susceptible to interference from other watermarks or image noise.

[0070] In some embodiments, the watermark preprocessing module also incorporates frequency domain compression techniques, such as quantization strategies, to further reduce the redundancy of the watermark information. Frequency domain compression can compress the volume of the watermark information, reduce the impact on image quality during embedding, and improve the transmission efficiency of the watermark.

[0071] In summary, the core function of the watermark preprocessing module is to perform orthogonal transformation and processing on the watermark image to ensure that the watermark information has high independence and robustness, effectively resisting image compression, noise, and other attacks. By selecting appropriate transformation methods, such as DCT, DFT, or SVD, the watermark preprocessing module in this embodiment can maximize the stability of the watermark during the embedding process and provide high-quality watermark information for the watermark embedding module, thereby ensuring the concealment and extraction accuracy of the final image watermark.

[0072] The watermark embedding module is used to embed the watermark into the low-frequency subband of the carrier image based on the output of the watermark preprocessing module, and adjust the embedding strength through singular value decomposition to match the watermark information with the features of the carrier image, thereby improving robustness and concealment.

[0073] In this embodiment, the main task of the watermark embedding module is to embed the watermark into the carrier image using appropriate technical means, based on the low-frequency sub-band (such as the LLLLLL sub-band) provided by the image preprocessing module and the processed watermark information provided by the watermark preprocessing module. Specifically, the watermark embedding module first performs singular value decomposition (SVD) on the low-frequency sub-band, then embeds the watermark information into the decomposed singular value matrix, and finally reconstructs the image using inverse SVD to generate the watermark-embedded carrier image.

[0074] Watermark embedding process

[0075] In this embodiment, the specific steps for watermark embedding are as follows:

[0076] First, the low-frequency subband of the image (i.e., obtained after DWT decomposition) is obtained from the image preprocessing module. Sub-bands are then processed and singular value decomposition is performed on them. Specifically, suppose the watermarked image is obtained after watermark preprocessing. Then for the low-frequency subband of the carrier image SVD processing yielded the following decomposition results:

[0077]

[0078] in, and These are the left singular matrix and the right singular matrix, which respectively contain the main feature information of the image; It is a singular value matrix that contains the energy information of the image.

[0079] Next, the watermark embedding module will orthogonally transform the watermark information. Embedded into the singular value matrix SSS. The embedding process uses the following formula:

[0080]

[0081] in:

[0082] It is the singular value matrix after embedding the watermark;

[0083] It is the first after preprocessing One watermark information;

[0084] It is a factor that controls the watermark strength and is usually optimized by variational methods to balance the amount of watermark information and the embedding strength, ensuring a balance between maximum information capacity and minimum visual impact.

[0085] In one possible implementation, embedding strength The watermark strength can be dynamically adjusted in different regions using optimization algorithms (such as genetic algorithms or gradient descent). Specifically, a stronger watermark is embedded in low-contrast areas of the image, while a weaker watermark is embedded in high-contrast areas.

[0086] As an alternative, to improve system stability and robustness, the watermark embedding module can also incorporate singular value compression technology to compress the embedded watermark information, thereby reducing redundancy and improving embedding efficiency. Specifically, during the embedding process, the singular value matrix... Some smaller singular values ​​in the image can be suppressed, thereby improving the concealment of the watermark and avoiding affecting image quality.

[0087] Embedded singular value matrix Then will be with and Together, they were reconstructed using Inverse Singular Value Decomposition (InverseSVD) to recover the new low-frequency subband. Its expression is:

[0088]

[0089] In some embodiments, the watermark embedding module can also be used for other high-frequency subbands. , , Local watermark embedding is performed to further increase the watermark's capacity and resistance to attacks. Although these high-frequency subbands are susceptible to compression and noise interference, proper embedding can improve the watermark's redundancy and recovery capability in specific scenarios. By applying different weights to each subband, the watermark information can be embedded with varying strengths, ensuring both image quality and watermark stability.

[0090] Generate watermarked images

[0091] After the watermark is embedded, the low-frequency subband after watermark embedding Will be in conjunction with other high-frequency subbands , , The images are then reassembled. The reassembled image is then reconstructed using inverse discrete wavelet transform (IDWT) to generate the final watermarked carrier image. The specific reassembly and reconstruction process is as follows:

[0092]

[0093] in, This results in the final watermarked image. During this process, the low-frequency sub-band... The embedded watermark information is not easily perceived, while the structural features of the image are well preserved.

[0094] Optimization and Adjustment

[0095] To further enhance the robustness and concealment of watermarks, the watermark embedding module can incorporate adaptive optimization based on image content. Specifically, the embedding module can pre-analyze the image using edge detection algorithms (such as the Canny or Sobel operators), selecting flat regions for watermark embedding to avoid impacting important structural parts of the image. Furthermore, by employing local embedding strength adjustment, the watermark strength can be dynamically adjusted based on the complexity and importance of the image content, resulting in a more uniform distribution of the watermark across the image and further improving concealment.

[0096] In this embodiment, the watermark embedding module performs singular value decomposition on the low-frequency subbands of the image and embeds the watermark information. Combined with optimized embedding strength and flexible application of multiple subbands, it achieves efficient embedding of watermark information into the image. Through inverse SVD reconstruction and inverse DWT reconstruction processes, a watermarked image is generated, ensuring image quality and watermark recoverability. Simultaneously, the watermark embedding module further improves the stability and robustness of the watermark through adaptive optimization techniques and local adjustments, ensuring reliable extraction of watermark information under various attacks such as compression, noise, and cropping.

[0097] The watermark optimization and enhancement module is used to optimize watermark embedding parameters. By dynamically adjusting the embedding strength, the watermark can maintain its concealment while improving information capacity and robustness, and reducing the similarity between watermarks.

[0098] In this embodiment, the watermark optimization and enhancement module employs variational methods and optimal control theory to adaptively adjust parameters during the watermark embedding process. Specifically, this module optimizes the embedding strength and position to maximize the amount of watermark information while minimizing the impact of the watermark on image quality. Through the optimized embedding scheme, the watermark can remain stably present in the image, and the watermark information can be reliably extracted even when facing common attacks such as compression, noise, and cropping.

[0099] Watermark optimization process

[0100] In this embodiment, the core objective of watermark optimization is to balance the watermark's concealment and robustness. Generally, the strength and position of the watermark embedding are influenced by the image content. Excessive watermark strength can affect image quality, while insufficient strength may prevent effective watermark extraction. Therefore, the watermark optimization module uses a variational method to optimize the watermark embedding strength, thereby ensuring that the watermark maximizes information storage in the image while maintaining good concealment.

[0101] In some embodiments, the watermark embedding module applies singular value matrices. After the embedding operation, the watermark optimization and enhancement module adjusts the watermark strength. The optimized embedding process is as follows:

[0102]

[0103] in:

[0104] It is the optimized singular value matrix, which contains watermark information;

[0105] For the preprocessed first One watermark information;

[0106] It is the strength factor of the watermark embedding, which needs to be dynamically adjusted through optimization algorithms.

[0107] The purpose of variational methods is to find the optimal... The goal of variational optimization is to maximize the capacity of watermark information while ensuring the watermark's concealment. Specifically, the objective of variational optimization is to minimize the similarity between the watermark and the image during the watermark embedding process. This process reduces the impact of watermark information on image quality by optimizing the adaptability of watermark strength to image content.

[0108] In some embodiments, the watermark optimization module can also adjust the recovery parameters during the embedding process using optimal control theory to further improve the watermark extraction accuracy. For example, in the watermark extraction stage, optimal control theory can help adjust the recovery parameters to ensure that errors are minimized during the extraction process. By dynamically adjusting the control parameters, the system can adapt to different image content and attack environments, thereby enhancing the robustness of the watermark.

[0109] Enhance watermark robustness

[0110] In this embodiment, the watermark optimization and enhancement module not only optimizes the watermark embedding strength but also further improves the robustness of the watermark under various attack conditions. Specifically, the watermark optimization and enhancement module enhances the image after watermark embedding to ensure that the watermark can still be stably restored when the image undergoes common attacks such as compression, noise, cropping, and rotation.

[0111] In some embodiments, the enhancement strategy includes the following methods:

[0112] (1) Adaptive adjustment of embedding strength

[0113] During the watermark embedding process, the watermark optimization and enhancement module adaptively adjusts the embedding strength to ensure that the watermark has appropriate strength in different areas of the image. In flat areas of the image, the watermark strength is greater to increase the watermark capacity; in detailed areas of the image, the watermark strength is lower to reduce the impact on the image structure.

[0114] Specifically, embedding strength The watermark can be adaptively adjusted in different regions of the image based on the image's gradient and texture information. In low-gradient regions (regions with small changes in the horizontal or vertical directions), the watermark can be embedded more strongly; while in high-gradient regions, the watermark intensity is reduced accordingly to avoid affecting the image details.

[0115] (2) Noise immunity and compression enhancement

[0116] To improve the watermark's robustness to noise, the watermark optimization and enhancement module employs denoising techniques such as Wiener filtering or median filtering. These filtering algorithms effectively remove noise and improve the stability of watermark information extraction. In compression environments, the watermark optimization module also adjusts the watermark embedding method to better adapt to changes in compression algorithms, ensuring that the watermark can still be effectively recovered after JPEG compression.

[0117] (3) Embedding redundant watermark information

[0118] To enhance system robustness, the watermark optimization and enhancement module can also employ redundant watermark information, embedding the same watermark information in multiple regions of the image. By redundantly storing the watermark information, the system can recover the watermark information from other regions even if some areas of the image are attacked or lost.

[0119] (4) Anti-rotation and anti-cutting reinforcement

[0120] The watermark optimization and enhancement module can also employ techniques such as Fourier-Mellin Transform (FMT) to enhance the robustness of the watermark against rotation and cropping attacks. Fourier-Mellin Transform ensures that the watermark information remains invariant to image rotation, guaranteeing accurate extraction even if the image is rotated.

[0121] In some embodiments, redundant watermark information can be distributed across multiple regions of the image using a spatial embedding strategy. This allows the system to recover the watermark information from other regions even if the image is cropped or rotated, improving the robustness of the watermark.

[0122] The watermark optimization and enhancement module maximizes the watermark embedding effect in images through a series of optimization algorithms and enhancement strategies. By dynamically adjusting the watermark strength, enhancing noise and compression resistance, introducing redundant information, and strengthening resistance to rotation and cropping, the module significantly improves the robustness and concealment of the watermark. During image processing, this module optimizes various parameters of watermark embedding to ensure the watermark information remains stable under various attacks and ensures accurate watermark extraction, ultimately improving the performance and reliability of the entire watermarking system.

[0123] The watermark extraction and restoration module is used to recover watermark information from images with embedded watermarks based on inverse discrete wavelet transform and inverse singular value decomposition. It also optimizes the extraction parameters based on optimal control theory, enabling it to adaptively optimize the extraction accuracy under different attack conditions and improve the stability of watermark restoration.

[0124] In this embodiment, the main operation steps of the watermark extraction and restoration module include inverse discrete wavelet transform (IDWT), inverse singular value decomposition (SVD), and extraction optimization based on optimal control theory. Through these techniques, the module can extract and restore watermark information from watermarked images, ensuring the stability and reliability of the watermark.

[0125] Watermark extraction process

[0126] In this embodiment, the watermark extraction process first extracts the low-frequency sub-band from the watermarked image and then performs Discrete Wavelet Transform (DWT) on it, decomposing the image into multiple frequency sub-bands. Specifically, the low-frequency sub-band is extracted by performing DWT processing on the watermarked image. Then, singular value decomposition (SVD) is performed to obtain the reconstructed singular value matrix. The mathematical representation of this process is as follows:

[0127]

[0128] in:

[0129] The low-frequency subband of the watermarked image;

[0130] and It is the corresponding singular matrix;

[0131] This is the singular value matrix recalculated after embedding the watermark.

[0132] In some embodiments, singular value matrices It may have been affected by the watermark information, therefore it needs to be compared with the singular value matrix in the original image. The embedded watermark information is extracted through comparison. This process can be performed using the following formula:

[0133]

[0134] in:

[0135] It is the first The extracted watermark information;

[0136] and These are the singular value matrices of the watermarked image and the original image, respectively.

[0137] It is the watermark strength coefficient, which needs to be restored based on the parameters when the watermark was embedded.

[0138] Watermark recovery

[0139] After watermark extraction, the restoration process uses inverse orthogonal transformation (such as inverse DCT or inverse DFT) to restore the extracted watermark information to the original watermark image. Specifically, the watermark information after orthogonal transformation... It can be recovered using the following formula:

[0140]

[0141] in:

[0142] It is the inverse transformation matrix of the orthogonal transformation;

[0143] The extracted watermark information;

[0144] This is the restored watermarked image.

[0145] In some embodiments, errors during the recovery process may affect the quality of the watermark. Therefore, this invention also introduces optimal control theory to dynamically adjust parameters during the recovery process. Using optimal control theory, the accuracy of watermark recovery can be optimized by minimizing the extraction error. In this process, the error in watermark recovery... It can be represented as:

[0146]

[0147] in:

[0148] EEE represents the recovery error;

[0149] This represents the Frobenius norm, used to measure the difference between the restored watermark and the original watermark;

[0150] The original watermark image. This is the extracted watermark image.

[0151] By using the optimal control algorithm, various parameters involved in the recovery process can be dynamically adjusted to minimize recovery errors and ensure accurate watermark recovery and stability.

[0152] Enhancing the robustness of watermark extraction

[0153] To improve the robustness of the watermark extraction process, this embodiment also introduces some optimization strategies. By combining noise filtering and image enhancement techniques, the impact of noise interference on watermark extraction can be effectively reduced. Under normal circumstances, images may be affected by noise during watermark extraction, especially during JPEG compression or other image compression processes, where noise can severely affect the recovery of watermark information. Therefore, this invention employs an adaptive filter (such as a Wiener filter) to denoise the image, thereby improving the accuracy of watermark information extraction.

[0154] Alternatively, the watermark extraction module can employ a redundancy coding strategy, embedding watermark information in multiple regions to increase the stability of watermark recovery. For example, when a portion of an image is cropped or damaged, redundancy coding allows for the recovery of complete watermark information from other regions even if some watermark information is lost.

[0155] Furthermore, regarding rotation resistance, the watermark extraction module enhances the watermark's robustness against rotation attacks through Fourier-Mellin transform. Specifically, the Fourier-Mellin transform of the watermark image ensures that the watermark information remains unchanged regardless of image rotation, allowing for accurate recovery of the watermark information even if the image is rotated.

[0156] The watermark extraction and restoration module employs a series of techniques to ensure the accurate extraction of watermark information from watermarked images. By combining inverse wavelet transform (IDWT), inverse singular value decomposition (SVD), and optimal control theory, watermark information can be efficiently restored. Simultaneously, the use of adaptive filtering and redundant watermark embedding strategies improves the system's robustness, enabling stable and accurate watermark restoration even after attacks such as compression, noise, cropping, and rotation. Ultimately, the watermark extraction and restoration module effectively guarantees accurate watermark restoration, enhancing the system's reliability under various environments.

[0157] The robustness enhancement module is used to improve the watermark extraction success rate under attacks such as JPEG compression, cropping, noise, and rotation. It enhances the watermark's resistance to cropping through a spatially distributed embedding strategy, enhances the watermark's resistance to rotation through Fourier-Mellin transform, and enhances the watermark's resistance to noise through an adaptive filtering strategy.

[0158] In this embodiment, the robustness enhancement module employs various techniques, such as redundant embedding, adaptive filtering, compression resistance enhancement, rotation resistance, and cropping enhancement, to enhance the stability of the watermark under different attacks. By combining these techniques, the robustness enhancement module ensures that the watermark maintains its effectiveness and recoverability under various image attacks.

[0159] Robustness enhancement process

[0160] In this embodiment, the robustness enhancement module ensures the stability of the watermark through various technical means. Specifically, the module's functions include the following aspects:

[0161] (1) Enhanced resistance to JPEG compression

[0162] Generally, JPEG compression is one of the main factors affecting watermark stability, especially at high compression ratios, where watermark information is often lost or becomes difficult to recover. To improve the ability to recover watermarks after compression, the robustness enhancement module first designs an adaptive quantization strategy for JPEG compression. During watermark embedding, the embedding position and strength of the watermark information are optimized based on the characteristics of the compressed image to ensure that the watermark information is effectively preserved.

[0163] Specifically, during watermark embedding, a quantization step is used to adjust the watermark intensity. At this stage, the low-frequency components of the image are typically less affected by compression, so the watermark is embedded more heavily in these areas. For higher-frequency regions, the watermark intensity is appropriately reduced to avoid excessive information loss. The quantized watermark information is adjusted using the following formula:

[0164]

[0165] in:

[0166] It is the quantified watermark information;

[0167] It embeds the previously displayed watermark information;

[0168] It is the quantization step size, which controls the watermark strength.

[0169] Through this process, the watermark can still maintain a certain degree of robustness after JPEG compression.

[0170] (2) Enhanced noise immunity

[0171] In practical applications, watermark extraction is often affected by noise, especially after the image has been transmitted or stored, where noise can significantly impact the accuracy of watermark recovery. To effectively reduce the impact of noise on watermark extraction, the robustness enhancement module employs adaptive filtering techniques, such as Wiener filtering and median filtering.

[0172] Wiener filtering is a commonly used denoising technique that removes noise from an image by calculating its local mean and variance. In watermark extraction, Wiener filtering can effectively remove high-frequency noise, ensuring stable extraction of the watermark information even under noise interference. Its mathematical expression is as follows:

[0173]

[0174] in:

[0175] This is the watermark information after noise reduction;

[0176] This is the original watermark information;

[0177] The noise standard deviation determines the strength of the filter.

[0178] Median filtering can remove salt-and-pepper noise from images, and in the process of watermark information recovery, combining median filtering can further enhance robustness.

[0179] (3) Resistance to shearing and rotation enhancement

[0180] In some attack scenarios, images may be subjected to cropping and rotation operations, which can lead to the loss of watermark information or its inability to be correctly recovered. To address this issue, the robustness enhancement module employs redundant embedding techniques and Fourier-Mellin transform (FMT).

[0181] Specifically, during the watermark embedding process, redundant embedding—that is, embedding the same watermark information simultaneously in multiple regions—enhances the watermark's resistance to cropping. For example, multiple regions of an image may store the same watermark information, so even if the image is cropped, the remaining watermark can still be completely recovered. The redundant distribution of watermark information can be represented by the following formula:

[0182]

[0183] in:

[0184] It is the watermark information after redundant embedding;

[0185] It is the weight of redundant information;

[0186] It is the first Watermark information for each region.

[0187] The Fourier-Mellin Transform (FMT) is a rotation-invariant transform that effectively handles changes in watermark information caused by image rotation. Through the Fourier-Mellin Transform, the watermark information of a rotated image is mapped to an invariant space, ensuring that the watermark information can still be recovered from the rotated image. Specifically, the Fourier-Mellin Transform can be calculated using the following formula:

[0188]

[0189] in:

[0190] This is the result of the watermark information after undergoing a Fourier-Mellin transform.

[0191] These are the pixel values ​​of the watermark image;

[0192] This is a rotation-invariant filter.

[0193] This method allows watermark information to be accurately recovered even if the image is rotated.

[0194] In this embodiment, the robustness enhancement module enhances the robustness of the watermark through various means, ensuring stable watermark information extraction even when faced with common attacks such as JPEG compression, noise, cropping, and rotation. Through techniques such as adaptive quantization, adaptive filtering, redundant embedding, and Fourier-Mellin transform, the robustness enhancement module significantly improves the system's resistance to various interferences. In practical applications, this module can significantly enhance the accuracy and stability of watermark extraction while maintaining image quality, thereby improving the security and reliability of the entire watermarking system.

[0195] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. An adaptive image watermark embedding and recovery system, characterized by, The system comprises: an image preprocessing module for performing discrete wavelet transform on a carrier image to be embedded with a watermark, decomposing the image into a plurality of frequency subbands, and selecting a low-frequency subband as a watermark embedding region; a watermark preprocessing module for performing orthogonal transform on a watermark image to maintain orthogonality among a plurality of watermarks and reduce mutual interference among the watermarks; a watermark embedding module for embedding the watermark into the low-frequency subband of the carrier image based on the output of the watermark preprocessing module, and adjusting embedding strength through singular value decomposition to match the watermark information with the carrier image features, improve robustness and concealment, and use the following formula in the embedding process: wherein: is the singular value matrix after embedding the watermark; is the pre-processed first watermark information; and is the pre-processed second watermark information. is a factor that controls the watermark intensity; a watermark optimization and enhancement module for optimizing watermark embedding parameters, dynamically adjusting embedding strength to improve information capacity and robustness while maintaining concealment after watermark embedding, and reducing similarity among watermarks; a watermark extraction and recovery module for recovering watermark information from the image embedded with the watermark based on inverse discrete wavelet transform and inverse singular value decomposition, and optimizing extraction parameters based on optimal control theory to adaptively optimize extraction accuracy under different attack conditions and improve watermark recovery stability; a robustness enhancement module for improving the success rate of watermark extraction under JPEG compression, cropping, noise, and rotation attacks, and enhancing watermark anti-cropping capability through spatial distributed embedding strategy, enhancing watermark anti-rotation capability through Fourier-Mellin transform, and enhancing watermark anti-noise capability through adaptive filtering strategy.

2. The adaptive image watermark embedding and recovery system of claim 1, wherein, The image preprocessing module performs discrete wavelet transform on the carrier image to generate a low-frequency subband, a high-frequency subband, and selects the low-frequency subband as a watermark embedding region to improve the robustness of the watermark.

3. The adaptive image watermark embedding and recovery system of claim 1, wherein, The watermark preprocessing module uses orthogonal transform to transform the watermark image, maintain orthogonality among a plurality of watermarks, and reduce mutual interference among the watermarks, wherein the orthogonal transform includes discrete cosine transform or discrete Fourier transform.

4. The adaptive image watermark embedding and recovery system of claim 1, wherein, The watermark embedding module decomposes the low-frequency subband of the carrier image through singular value decomposition, and embeds the watermark information into the singular value matrix obtained by decomposition. The singular value matrix after embedding is used to reconstruct the low-frequency subband, and the image with watermark is reconstructed through inverse discrete wavelet transform.

5. The adaptive image watermark embedding and recovery system of claim 1, wherein, The watermark optimization and enhancement module optimizes watermark embedding strength based on variational method, and dynamically adjusts embedding parameters through optimization algorithm to improve information capacity and robustness while maintaining concealment after watermark embedding.

6. The adaptive image watermark embedding and recovery system of claim 1, wherein, The watermark extraction and recovery module performs discrete wavelet transform on the image with watermark, extracts the low-frequency subband, calculates the embedded watermark information based on singular value decomposition, and then recovers the original watermark using inverse orthogonal transform.

7. The adaptive image watermark embedding and recovery system of claim 1, wherein, The watermark extraction and recovery module uses optimal control theory to dynamically adjust watermark recovery parameters to adaptively optimize extraction accuracy under different attack conditions and improve watermark recovery stability.

8. The adaptive image watermark embedding and recovery system of claim 1, wherein, The robustness enhancement module uses a quantization adaptive strategy to adjust the quantization step of watermark embedding to make the watermark information still recoverable after JPEG compression.

9. The adaptive image watermark embedding and recovery system of claim 1, wherein, The robustness enhancement module stores the watermark information in multiple regions through a spatial distributed embedding strategy to enhance the resistance of the watermark to cropping attacks.

10. The adaptive image watermark embedding and recovery system of claim 1, wherein, The robustness enhancement module adopts Fourier-Mellin transform to transform the embedded watermark information to enhance the robustness of the watermark to rotation attacks. For noise attacks, the robustness enhancement module adopts an adaptive filtering strategy including Wiener filtering and median filtering to reduce the influence of noise on watermark extraction and combines redundancy coding to improve the stability of watermark extraction.

Images