Multi-position overlapped image privacy protection method and system based on compressed sensing

Abstract

The invention provides a multi-position overlapping image privacy protection method and system based on compressed sensing, and relates to the technical field of compressed sensing and image privacy protection, and the method comprises the steps: obtaining an original image, and dividing privacy parts and non-privacy parts of different parts of the original image; generating a coincident privacy mask for the coincident privacy part, generating a privacy mask for the non-coincident privacy part, confusing all the privacy parts, and performing image reconstruction by using the confused privacy part and the original non-privacy part; sampling encryption of compressed sensing is carried out on the reconstructed image, a compressed sensing ciphertext is generated, corresponding watermark matrixes and watermark embedding matrixes are generated for different privacy parts according to a confusion matrix, the watermark embedding matrixes and the watermark matrixes are embedded into the compressed sensing ciphertext, and encryption of the compressed sensing ciphertext is achieved; and users with different levels of keys are utilized to decode and reconstruct different parts of images of the encrypted compressed sensing ciphertext, so that privacy protection of multi-position overlapped images is realized.

Description

Technical Field

[0001] The present disclosure relates to the technical field of compressed sensing and image privacy protection, and in particular to a method and system for privacy protection of multi-position overlapping images based on compressed sensing. Background Art

[0002] The statements in this section merely provide background information related to the present disclosure and do not necessarily constitute prior art.

[0003] With the rapid development of information technology and digital image processing technology, the acquisition, transmission and sharing of image data have become extremely convenient. In the fields of security monitoring, intelligent transportation, social media, etc., the application of large amounts of image data has improved people's quality of life and safety. However, the widespread dissemination of image data also brings the risk of personal privacy leakage.

[0004] Compressed sensing is an emerging signal processing technology that can acquire and accurately reconstruct signals at a rate far lower than the Nyquist sampling rate during the signal sampling stage. Applying compressed sensing to the field of image privacy protection can protect sensitive information in the image while reducing the amount of data, and improve processing and transmission efficiency. At the same time, the advantage of compressed sensing is that as long as the original signal can be transformed in a sparse orthogonal transform domain, it can effectively recover a large amount of data with fewer sample measurements.

[0005] In recent years, various encryption and privacy protection schemes based on compressed sensing theory have been proposed and widely used in various fields. Compared with traditional full-image encryption and compression methods, privacy protection schemes based on compressed sensing can reduce computational complexity and meet the needs of real-time processing. At the same time, the random measurement process of compressed sensing has encryption characteristics, and it is difficult for unauthorized users to reconstruct the original image from compressed data, which effectively protects privacy information and improves data transmission and storage efficiency.

[0006] However, existing compressed sensing-based image privacy protection methods mainly focus on global or single-region privacy protection, or privacy protection for independent parts of multiple locations. There is still a lack of effective solutions for privacy areas in multiple locations with overlapping parts. These methods may not be able to balance privacy protection and image quality when facing complex scenes, which limits their promotion in practical applications. In addition, under the compressed sensing framework, how to efficiently embed and extract watermarks in the image encryption process to ensure the robustness and security of the watermarks without affecting the reconstruction quality of the image is an urgent problem to be solved. Summary of the invention

[0007] In order to solve the above problems, the present invention proposes a multi-position overlapping image privacy protection method and system based on compressed sensing. In the image encryption process, the watermark embedding and extraction methods are improved to make it more robust and secure under the compressed sensing framework, thereby not affecting the reconstruction quality of the image; strengthen the protection of personal sensitive information and improve the transmission and storage efficiency of data.

[0008] According to some embodiments, the present disclosure adopts the following technical solutions: The privacy protection method of multi-position overlapping images based on compressed sensing includes: Acquire the original image, pre-process the original image, and divide the privacy part and the non-privacy part of different parts of the original image; Locate the overlapping private parts, generate overlapping privacy masks for the overlapping private parts, generate privacy masks for the non-overlapping private parts, use the generated overlapping privacy masks and privacy masks as confusion matrices to confuse all private parts, and reconstruct the image using the confused private parts and the original non-private parts to obtain a reconstructed image; Performing compressed sensing sampling encryption on the reconstructed image to generate compressed sensing ciphertext, generating corresponding watermark matrices and watermark embedding matrices for different privacy parts according to the confusion matrix, embedding the watermark embedding matrix and the watermark matrix into the compressed sensing ciphertext, and realizing encryption of the compressed sensing ciphertext; Users with different levels of keys can decode and reconstruct different parts of the encrypted compressed sensing ciphertext to achieve privacy protection of overlapping images in multiple locations.

[0009] According to some embodiments, the present disclosure adopts the following technical solutions: The multi-position overlapping image privacy protection system based on compressed sensing includes: A data acquisition module, used to acquire the original image, pre-process the original image, and divide the privacy part and the non-privacy part of different parts of the original image; An image reconstruction module is used to locate the overlapping private parts, generate an overlapping privacy mask for the overlapping private parts, generate a privacy mask for the non-overlapping private parts, use the generated overlapping privacy mask and privacy mask as a confusion matrix to confuse all the private parts, and use the confused private parts and the original non-private parts to reconstruct the image to obtain a reconstructed image; A watermark embedding module is used to perform compressed sensing sampling encryption on the reconstructed image to generate compressed sensing ciphertext, generate corresponding watermark matrices and watermark embedding matrices for different privacy parts according to the confusion matrix, embed the watermark embedding matrix and the watermark matrix into the compressed sensing ciphertext, and realize encryption of the compressed sensing ciphertext; The decoding and reconstruction module is used to decode and reconstruct different parts of the encrypted compressed sensing ciphertext using users with different levels of keys, thereby realizing the privacy protection of overlapping images in multiple locations.

[0010] According to some embodiments, the present disclosure adopts the following technical solutions: A computer program product includes a computer program, and when the computer program is executed by a processor, the multi-position overlapping image privacy protection method based on compressed sensing is implemented.

[0011] According to some embodiments, the present disclosure adopts the following technical solutions: A non-transitory computer-readable storage medium is used to store computer instructions. When the computer instructions are executed by a processor, the multi-position overlapping image privacy protection method based on compressed sensing is implemented.

[0012] According to some embodiments, the present disclosure adopts the following technical solutions: An electronic device comprises: a processor, a memory and a computer program; wherein the processor is connected to the memory, the computer program is stored in the memory, and when the electronic device is running, the processor executes the computer program stored in the memory, so that the electronic device implements the multi-position overlapping image privacy protection method based on compressed sensing.

[0013] Compared with the prior art, the present invention has the following beneficial effects: The disclosed multi-position overlapping image privacy protection method based on compressed sensing generates independent overlapping privacy masks for privacy areas at multiple locations with overlapping parts by accurately locating the privacy areas, and performs special processing on the overlapping areas. During the encoding process, a joint obfuscation and encryption strategy is adopted for the overlapping privacy areas to ensure that all privacy information is effectively protected. Compared with traditional methods, the disclosed method can avoid privacy leakage caused by overlapping areas and improve the comprehensiveness and reliability of privacy protection.

[0014] The disclosed method for privacy protection of multi-position overlapping images based on compressed sensing optimizes the embedding and extraction methods of watermarks under the compressed sensing framework. By embedding watermark information in the compressed domain, the influence of the watermark on the original image is reduced, and the robustness and invisibility of the watermark are improved. The watermark information is not easily lost during the image compression, transmission and reconstruction process, which ensures the copyright and integrity of the image. Compared with the traditional watermark method, the watermark processing disclosed in the present invention is more suitable for the environment of compressed sensing encryption, which enhances the security and effectiveness of the watermark.

[0015] The privacy protection method for multi-position overlapping images based on compressed sensing disclosed in the present invention realizes real-time processing of images by optimizing the computational complexity of the algorithm and utilizing parallel computing and fast reconstruction technology. The compressed sensing encryption is directly performed on the image during the sampling stage, which reduces the amount of data and improves the transmission and storage efficiency. Compared with the traditional full image encryption and high-resolution image processing methods, the present invention can significantly improve the processing speed and efficiency while ensuring security. The method disclosed in this disclosure is applicable to various complex scenarios and can adaptively adjust sampling and encryption strategies according to different privacy requirements and image features. Whether it is a single privacy area, multiple non-overlapping privacy areas, or multiple overlapping privacy areas, it can be effectively protected and processed. Compared with the limitations of the prior art, the method disclosed in this disclosure has better applicability and versatility and has broad application prospects. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] The accompanying drawings constituting a part of the present disclosure are used to provide a further understanding of the present disclosure. The illustrative embodiments of the present disclosure and their descriptions are used to explain the present disclosure and do not constitute an improper limitation on the present disclosure.

[0017] Figure 1 is a flowchart of the overall architecture of an embodiment of the present disclosure; Figure 2 Schematic diagram of reconstruction quality of different images according to an embodiment of the present disclosure; Figure 3 is a histogram correlation test result of the original image dog according to an embodiment of the present disclosure; Figure 4 is a histogram correlation test result of the original image sheep in the embodiment of the present disclosure; Figure 5 is a histogram correlation test result of the original image cat according to an embodiment of the present disclosure; Figure 6 is a histogram correlation test result of the encrypted image dog according to an embodiment of the present disclosure; Figure 7 is a histogram correlation test result of the encrypted image sheep according to an embodiment of the present disclosure; Figure 8 is a histogram correlation test result of the encrypted image cat according to an embodiment of the present disclosure; Fig. 9 Schematic diagram of the correlation analysis results of adjacent pixels of the original image dog at horizontal, vertical and diagonal positions according to an embodiment of the present disclosure; Fig.10 It is a schematic diagram of the correlation analysis results of adjacent pixels of the original image sheep at horizontal, vertical and diagonal positions according to an embodiment of the present disclosure; Fig.11Schematic diagram of the correlation analysis results of adjacent pixels of the original image cat at horizontal, vertical and diagonal positions according to an embodiment of the present disclosure; Fig.12 It is a schematic diagram of the correlation analysis results of adjacent pixels of the encrypted image dog at horizontal, vertical and diagonal positions according to an embodiment of the present disclosure; Fig.13 It is a schematic diagram of the correlation analysis results of adjacent pixels of the encrypted image sheep at horizontal, vertical and diagonal positions according to an embodiment of the present disclosure; Fig.14 It is a schematic diagram of the results of correlation analysis of adjacent pixels of the encrypted image cat at horizontal, vertical and diagonal positions according to an embodiment of the present disclosure. DETAILED DESCRIPTION

[0018] The present disclosure is further described below in conjunction with the accompanying drawings and embodiments.

[0019] It should be noted that the following detailed descriptions are all illustrative and are intended to provide further explanation of the present disclosure. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art to which the present disclosure belongs.

[0020] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit the exemplary embodiments according to the present disclosure. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "comprising" and / or "including" are used in this specification, it indicates the presence of features, steps, operations, devices, components and / or combinations thereof.

[0021] Example 1 In one embodiment of the present disclosure, a method for privacy protection of multi-position overlapping images based on compressed sensing is provided, comprising: Step 1: Obtain the original image, pre-process the original image, and divide the privacy part and non-privacy part of different parts of the original image; Step 2: Locate the overlapping private parts, generate overlapping privacy masks for the overlapping private parts, generate privacy masks for the non-overlapping private parts, use the generated overlapping privacy masks and privacy masks as confusion matrices to confuse all private parts, and use the confused private parts and the original non-private parts to reconstruct the image to obtain a reconstructed image; Step 3: Performing compressed sensing sampling encryption on the reconstructed image to generate compressed sensing ciphertext, generating corresponding watermark matrices and watermark embedding matrices for different privacy parts according to the confusion matrix, embedding the watermark embedding matrix and the watermark matrix into the compressed sensing ciphertext, and realizing encryption of the compressed sensing ciphertext; Step 4: Users with different levels of keys decode and reconstruct different parts of the encrypted compressed sensing ciphertext to achieve privacy protection of overlapping images in multiple locations.

[0022] As an embodiment, the specific implementation process of the multi-position overlapping image privacy protection method based on compressed sensing disclosed in the present invention is as follows: Step 1: Obtain the original image, pre-process the original image, and divide the privacy part and non-privacy part of different parts of the original image; Specifically, an original image to be processed is input, the size of the original image is obtained, and the original image is preprocessed to divide the original image into different parts of private and non-private parts, which are represented by different matrices, wherein the private positions of the matrix of the non-private part are all zero, and the matrix of the private part, except for the private position, the rest of the positions are all zero; Set the position element of the private part to 1 and the rest to zero to obtain the privacy masks of different private parts; Define the non-private part of the image signal and the private parts at different positions. First, determine whether the areas of the private parts overlap. If the positions of two private areas overlap or are close to overlapping, the overlapping parts need to be specially processed during the encoding (encryption) and decoding (decryption) processes to ensure the complete protection and correct recovery of the privacy information.

[0023] That is, the location information of multiple private parts is obtained, and the location coordinates of the upper left corner and the lower right corner of the private parts are recorded, converted into binary, and the converted binary values ​​are binarized accordingly so that the final value is 1 or -1, and the number of generated element bits is p.

[0024] Step 2: Locate the overlapping private parts, generate overlapping privacy masks for the overlapping private parts, generate privacy masks for the non-overlapping private parts, use the generated overlapping privacy masks and privacy masks as confusion matrices to confuse all private parts, and use the confused private parts and the original non-private parts to reconstruct the image to obtain a reconstructed image; Define the non-privacy part of the image signal and the privacy part at different positions. First, determine whether the area of ​​the privacy part overlaps. If the positions of the two privacy areas overlap or are close to overlapping, the overlapping parts need to be specially processed during the encoding (encryption) and decoding (decryption) process to ensure the complete protection and correct recovery of the privacy information. In the process of image recovery, in the processing of the first type of decoder, the overlap of the privacy part will not affect the reconstruction of ordinary users, so the reconstruction of the non-privacy part will still be reconstructed according to the privacy protection scheme of multiple privacy positions; for the second type of decoder (administrator users who manage the level I privacy area), there is no need to elaborate too much in the process of restoring each privacy position separately, which is basically consistent with the existing scheme; however, in the processing of the third type of decoder (advanced authorized users manage the privacy area of ​​the overlapping part), it is necessary to clarify the overlapping position of the privacy area position and perform corresponding processing on the overlapping position, so as to accurately reconstruct the complete image data. Based on the above, the present invention assigns a unique identifier to each privacy area. If there is an overlap between the two parts, the overlapping area is marked as over.

[0025] Specifically, the overlapping private parts include the Level II private parts and the Level I private parts, and the Level I private parts are nested in the Level II private parts; Generate a matrix of the same size as the original image, with randomly distributed -1 and 1 values ​​in the matrix elements. Multiply the masks corresponding to multiple private positions by the generated randomly distributed -1 and 1 matrices to obtain the corresponding confusion matrix. At this time, only the position information of the corresponding private parts in the confusion matrix is ​​-1 or 1, and the rest of the positions are all zero.

[0026] When processing the private part of a single position, a corresponding binary mask is generated as a privacy mask according to the marked private parts. For the overlapping private parts, a new mask (mask C) is separately marked to represent the overlapping parts, and the mask of the overlapping parts corresponding to each privacy area is replaced with the new mask as the overlapping privacy mask.

[0027] The generated overlapping privacy masks and privacy masks are used as confusion matrices to confuse all private parts to obtain the information of the confused private positions. The information of the confused private parts is then added to the information of the non-sensitive parts to obtain the reconstructed image. The entire reconstructed image is then sampled and encrypted using compressed sensing.

[0028] Step 3: Generate corresponding watermark matrices and watermark embedding matrices for different privacy parts according to the confusion matrix, and embed the watermark embedding matrix and the watermark matrix into the compressed sensing ciphertext, including: Remove the non-zero rows in the generated overlap privacy mask and privacy mask matrix to obtain two new matrices. Invert all the masks except the position of privacy region I in the overlap privacy mask corresponding to the overlap privacy region II. The transformed matrix is ​​called the initial watermark matrix, and the positions of the continuous zero elements in the initial watermark matrix are replaced with the processed elements of the binary code marked with the partial information of different privacy regions to form a watermark matrix. According to the size of the watermark matrix, the corresponding watermark embedding matrix and left zeroing matrix are generated. The watermark embedding matrix of the overlap privacy part II is multiplied with the watermark matrix, and the obtained data is added to the compressed sensing ciphertext of the compressed sensing to generate the final encrypted signal, and then the perturbation matrix is ​​added to the encrypted signal. In order to ensure the security of the signal, the scrambling encryption operation is applied to make it able to resist statistical analysis attacks. The perturbation matrix is ​​added to the encrypted signal to make the information distribution on the encrypted signal as uniform as possible.

[0029] The formation of the watermark matrix is ​​as follows: the non-zero rows of the confusion matrix are extracted to generate a new matrix, which is the initial watermark matrix. The position information of the private part after binary processing is embedded into the first p positions of the initial watermark matrix to generate the final watermark matrix.

[0030] Step 4: Users with different levels of keys decode and reconstruct different parts of the encrypted compressed sensing ciphertext to achieve privacy protection of overlapping images in multiple locations.

[0031] Specifically, at the receiving end, users with different levels of keys (ordinary users, administrator users, and advanced users) can use their respective keys to reconstruct different parts of the image, and the encrypted signals received by all users are the same. Each authorized user has the key of the scrambled matrix of the second layer of encryption. The first operation of each authorized user is to restore the second layer of encrypted signal with the key of the scrambled matrix. Then, the subsequent decoding operations corresponding to users with different permissions are carried out.

[0032] Ordinary users only have compressed sensing encryption keys. Such authorized users treat the added watermark part as noise and then use the reconstruction algorithm to obtain an estimate of the image data of the non-private part.

[0033] The administrator user has the compressed sensing encryption key and the intermediate watermark key B1, so as to restore the image data that the user can obtain, and this part of the image data can restore the information of one of the private parts. First, the watermark part is ignored as noise, and the reconstruction algorithm is used to obtain the original estimate of the sparse signal; the watermark information corresponding to the private part is eliminated by the left zeroing factor corresponding to the watermark, and the remaining information is treated as noise for corresponding processing, and the reconstruction algorithm is used to obtain the estimate of the original signal after the noise is removed; the compressed sensing encrypted signal is subtracted from the product of the measurement matrix and the original signal obtained by the operation to obtain the information of the embedded watermark part; the obtained watermark is embedded with the key B1 to obtain the watermark matrix; according to the corresponding information of the elements in the watermark matrix, according to the element encoding rules in the watermark matrix, the position information of the private part and the mask of the corresponding position are restored; according to the restored mask, the element information of the private position is restored, and the information of the estimated private part of the original signal is replaced with the restored element information of the private position to obtain the final restored data, and finally the image information of a private part and a non-private part is reconstructed.

[0034] Advanced users have compressed sensing encryption keys and higher-level watermark key B, and can recover all the information of the image. First, the watermark part is also regarded as additional noise, and the sparse signal is estimated by the reconstruction algorithm. Then, it is left-multiplied by a left zeroing factor to eliminate the noise of the watermark part corresponding to the advanced authorized user. Then, the product of the left zeroing matrix and the measurement matrix is ​​used as the new measurement matrix to obtain the accurately reconstructed original information using the reconstruction algorithm; the received encrypted signal is subtracted from the product of the measurement matrix and the estimated original signal obtained by the operation to obtain the information of the embedded watermark part; the watermark part information is restored with key B to obtain the watermark matrix corresponding to the private part; according to the restored watermark matrix, the mask of the corresponding position of the private part is restored according to the element encoding rule in the watermark matrix and the position information corresponding to the private part is obtained; according to the restored mask, the accurately reconstructed position information of the private part is replaced with the restored element information of the private part to obtain the restored original image information, and finally all the information of the original image is successfully reconstructed.

[0035] As an embodiment, the present disclosure takes two overlapping privacy-sensitive parts as an example, and the specific process of implementing the multi-position overlapping image privacy protection method based on compressed sensing of the present disclosure is as follows: 1: Obtain the original image signal, pre-process the original image signal, and divide the privacy part and the non-privacy part of different parts of the original image signal; The original image signal is a sparse or sparse signal , its linear compression measurement It can be expressed as:

[0036] in, is a sparse basis, , s contains The sparse coefficient of is the measurement matrix of M≪N, For dictionary.

[0037] Furthermore, the measurement matrix The adaptive compressed sensing measurement matrix introduced in this disclosure dynamically adjusts the parameters of the measurement matrix according to the distribution and characteristics of the privacy areas. For overlapping privacy areas, a dedicated measurement matrix is ​​designed. , improving encryption strength and reconstruction accuracy. This method can optimize the performance of compressed sensing and improve the quality of image reconstruction while ensuring privacy protection. In this disclosure, the measurement matrix is ​​adjusted according to the sparsity of the region. For regions with higher sparsity, the measurement matrix It can be designed to be more sparse. For privacy areas, increase the randomness of the measurement matrix to enhance encryption strength. When generating, the features of each privacy area involved in the overlapping area are fused to obtain the comprehensive features of the overlapping area, and then the measurement matrix is ​​customized according to the real features. , ensuring that the image information can be effectively reconstructed. The measurement matrix design can be flexibly adjusted according to actual needs and scenarios to achieve personalized privacy protection strategies.

[0038] 2: Reconstruct the code using the knowledge of 2-D compressed sensing; The original image signal X is classified into three categories: level I privacy parts, level II privacy parts, and non-privacy parts. Level I privacy parts are embedded in level II privacy parts, also known as advanced privacy parts. This privacy part is only accessible to users with the highest level of authority. Other ordinary users and authorized users (administrators) of level I privacy parts have no right to view the privacy content of this part, thereby ensuring the privacy management of users with higher authority (advanced users):

[0039] Because the high-level authorized user contains the information of the low-level authorized user, be careful not to repeat the operation when encoding to avoid problems. In the classified privacy part, the size of the matrix is ​​related to the original image signal. The pixel size is the same, all N×N, non-private parts is a matrix with all zeros in the private part. On the contrary, the private part , is a matrix whose corresponding private positions are non-zero and whose non-private positions are all zero.

[0040] Furthermore, we first need to perform preliminary obfuscation on the original image signal matrix, and then perform operations on the private parts. First, we generate two different binary masks, generate overlapping privacy masks for the overlapping private parts, and generate privacy masks for the non-overlapping private parts. It should be noted that the elements of the two binary masks for the corresponding positions of the level I privacy parts must be the same, and the elements of the remaining positions of the level II privacy parts except the positions corresponding to the level I privacy parts cannot be the same, otherwise, when performing the recovery operation, the level I privacy parts will bypass the level II privacy parts to restore them, so that the security of users with different permissions cannot be guaranteed accordingly.

[0041] After generating the corresponding binary mask, the binary mask is confused with the corresponding privacy parts of different levels to obtain the preliminary confusion matrix M1 and M2. The obfuscated privacy parts are combined with the non-privacy parts to perform compressed sensing sampling encryption to obtain the encrypted compressed sensing ciphertext signal Y e :

[0042] 3: Generate corresponding watermark matrix and watermark embedding matrix for different privacy parts according to the confusion matrix, embed the watermark embedding matrix and watermark matrix into the compressed sensing ciphertext to realize the encryption of the compressed sensing ciphertext; According to the generated confusion matrix, the corresponding watermark matrix is ​​generated. The generation of the watermark matrix mainly selects the non-zero rows corresponding to the binary mask to form the corresponding watermark matrix. In the watermark matrix, the encoding of the location information of the privacy areas corresponding to different privacy parts is also placed; according to the watermark matrix, the corresponding watermark embedding matrix is ​​generated. Because it involves overlapping parts, it is only necessary to generate the corresponding watermark embedding matrix, and then the watermark embedding matrix of the level II privacy part is multiplied by the watermark matrix and added to the compressed sensing ciphertext to obtain the ciphertext Y w , where the watermark embedding matrix for the level II privacy part is B, and the watermark embedding matrix for the level I privacy part is B1 ( ), the generated watermark matrices are Delt (Level II privacy part) and Delt1 (Level I privacy part), where , the corresponding formula is:

[0043] In order to enhance the security of the signal and avoid illegal attacks, an additional operation is added to the final ciphertext: the scrambling matrix Add is used to further scramble the ciphertext, so that the element information on the ciphertext is distributed more evenly, enhancing the security of the information during transmission and transportation. After another scrambling, the final information is represented as Y ec :

[0044] 4: Users with different levels of keys can decode and reconstruct different parts of the encrypted compressed sensing ciphertext to achieve privacy protection of overlapping images in multiple locations; After the encoding operation is completed, the signal is transmitted or transported to the receiver, and different authorized users receive the same encrypted signal. , but the keys owned by users of different levels are different: ordinary users have the scrambling key Add and the compressed sensing key , can use the key to reconstruct the image information of non-sensitive parts. The reconstruction algorithm we use here is the improved SPL algorithm. The reconstruction algorithm of other authorized users will also use this algorithm in the future. I will not go into details later. The administrator user has the scrambling key Add and the compressed sensing key and the key B1 of the level I privacy part, which can restore the image of the level I privacy part and the image of the non-privacy part; the advanced authorized user has the scrambling key Add, the compressed sensing key And the key B2 of the level II privacy part can restore all the image information. Among the three types of authorized users involved in this disclosure, the recovery operations of the administrator user and the senior authorized user can be performed separately, without nested operations, which effectively guarantees the independence of the decoding process of authorized users at all levels. The operations of the three types of authorized users are given below: Ordinary users only have the scrambling key Add and the compressed sensing key , first use the scrambling key to restore the received signal to the ciphertext part Y w , and then use the compressed sensing key to obtain the ciphertext During the recovery process, the additional watermark part is regarded as the generated noise, and then Minimize the scheme to restore the sparse signal to get an estimate of the original signal. The specific algorithm is as follows Table 1: Table 1 Common user decoding algorithm

[0045] The administrator user has the scrambling key Add and the compressed sensing key and watermark key B1, first use the scrambling key to recover the ciphertext, then multiply the ciphertext with the left zeroing matrix F1 to eliminate the corresponding watermark part, and regard the rest as noise, then use the left zeroing matrix and the measurement matrix key The matrix obtained by multiplication is regarded as the new measurement matrix, the initial value of X is estimated, and then the measurement matrix key is subtracted from the ciphertext The product of the estimated value can restore the watermark of the level I privacy area, and restore the elements of the privacy area according to the watermark, so as to reconstruct the non-privacy area and the level I privacy area. The specific algorithm is as follows in Table 2: Table 2 Administrator user decoding algorithm

[0046] Advanced authorized users have the scrambling key Add and the compressed sensing key and watermark key B, first use the scrambling key to recover the ciphertext, then multiply the ciphertext with the left zeroing matrix F to eliminate the corresponding watermark part, and regard the rest as noise, then combine the left zeroing matrix with the measurement matrix key The matrix obtained by multiplication is regarded as the new measurement matrix, the initial value of X is estimated, and then the measurement matrix key is subtracted from the ciphertext The product of the estimated value can restore the watermark of the level II privacy part, and restore the elements of the privacy part according to the watermark, so as to restore the entire image data. The specific algorithm is as follows in Table 3: Table 3 Advanced authorized user decoding algorithm

[0047] The present disclosure proposes a privacy protection mechanism for different permission management of overlapping parts, and processes the corresponding watermarks in different ways. In order to ensure the security of the ciphertext, the present disclosure also adds an interference mechanism to the ciphertext - scrambling key to ensure that it is not attacked maliciously. The scheme proposed in the present disclosure fully utilizes the encryption and sampling performance of compressed sensing, adopts multiple encryption to ensure the security and reliability of data, and can be effectively applied to industries involving privacy protection. The reconstruction algorithm of the present disclosure adopts the improved-SPL reconstruction algorithm, which effectively guarantees the reconstruction quality of the image at a low sampling rate.

[0048] Figure 2 The method of this embodiment is to restore frame performance of different authorized users in different images under the same sampling rate (MR) in image data involving multiple overlapping private parts.

[0049] Table 4 shows the change of the peak signal-to-noise ratio (PSNR) of the image under different sampling rates. It can be seen intuitively that as the sampling rate increases, the PSNR of the reconstructed image also increases, and the quality of image restoration also improves.

[0050] Table 4 Changes in peak signal-to-noise ratio (PSNR) of images under different sampling rates

[0051] Table 5 shows the structural similarity (SSIM) of images at different sampling rates. The results show that the SSIM values ​​of ordinary users are generally low, while the SSIM values ​​of administrator-authorized users and advanced-authorized users are relatively improved, and the value of advanced-authorized users is the best, especially when the sampling rate is higher, the SSIM value is higher. This accurately shows that the reconstruction quality of the image maintains a high structural similarity.

[0052] Table 5 Structural similarity (SSIM) of images with different sampling rates

[0053] The randomness of elements in ciphertext is one of the important indicators for evaluating the security of privacy protection schemes. The more random the element distribution is, the more difficult it is for the encrypted data to be cracked by attackers through statistical means, thereby improving the security of the scheme. By analyzing the histograms of the original image and ciphertext, it can be determined whether the proposed privacy protection method can resist statistical analysis attacks.

[0054] Simulation test This disclosure conducts a histogram test on a set of images containing overlapping privacy areas at multiple locations. The results are as follows: Figure 3 As shown. Figure 3 It can be seen that the histogram distribution of the ciphertext after processing by the present disclosure is close to uniform random distribution, the frequency of each grayscale pixel value is roughly the same, and there is a lack of obvious statistical characteristics. Compared with the histogram of the original image, the histogram of the ciphertext no longer has statistical information that can be utilized. This uniform histogram distribution shows that the ciphertext has eliminated the statistical characteristics of the original image, and it is difficult for attackers to obtain useful information through statistical analysis methods. Therefore, the privacy protection scheme of the present disclosure has significant advantages in resisting statistical analysis attacks, enhances the security and reliability of data, and provides strong protection for users' privacy data.

[0055] Another important aspect of evaluating the security of encrypted image systems is the correlation between adjacent pixels. In natural images, adjacent pixels usually have high correlation in the horizontal, vertical and diagonal directions. Attackers may use this high correlation to extract useful information from the image through statistical analysis, thereby destroying the security of encryption. Therefore, an effective image encryption scheme should be able to reduce the correlation between adjacent pixels in the ciphertext to resist such attacks. To this end, the correlation of adjacent pixels was tested on the original image and the encrypted image. The experimental results are shown in Figure 4. It can be observed that the distribution of adjacent pixel values ​​in the original image shows obvious regularity and shows strong correlation. In the encrypted image, the distribution of adjacent pixel values ​​tends to be randomized and the correlation is significantly reduced. The encryption scheme of the present invention successfully destroys the high correlation between adjacent pixels and prevents attackers from obtaining valuable information through statistical means. Therefore, the scheme provided by the present disclosure has obvious advantages in security, can effectively resist attacks based on statistical analysis, and protect the privacy and data security of users.

[0056] In order to further evaluate the security of the encryption scheme disclosed in the present invention, the information entropy analysis of the encrypted image was performed. Information entropy is an important indicator to measure the randomness of the grayscale value distribution of image pixels. The higher the entropy value, the more random the distribution of pixel values, the stronger the unpredictability of the image, and the higher the security. In the experiment, the information entropy of multiple encrypted images was calculated, and the results are summarized in Table 6. The results show that the information entropy of these encrypted images is close to 8, which is the maximum entropy value that can be achieved by 8-bit grayscale images. This means that the distribution of pixel values ​​of the encrypted image is very close to completely random, and it is difficult for attackers to obtain valuable information from it through statistical methods. In addition, the high information entropy of the ciphertext image also shows that the encryption process disclosed in the present invention effectively eliminates the statistical characteristics of the original image and enhances the irreversibility and unpredictability of the encrypted image. In summary, from the results of the information entropy analysis, the encryption scheme disclosed in the present invention has significant advantages in terms of security.

[0057] Table 6 Entropy analysis of different secret-detecting images

[0058] Among the binary masks generated by the two private parts, since the overlapping private positions are involved, the level II private part contains the mask of the level I private position, so the binary mask of the level II private part will be processed. The image pixels used in the public experiment are all 512×512 in size, so the size of the corresponding generated binary mask is the same as the original image size. Since the mask of the level II private part is sparse, only the non-zero rows of this part of the mask are selected to form an initial watermark matrix to know the mask of the private part I; the non-zero rows of the mask corresponding to the obtained privacy area II are called the initial watermark W, and the non-zero rows of the mask matrix generated by the private part I are The first row is called the initial watermark w1. All non-zero elements in W except the privacy area I are inverted, and the corresponding initial watermark is disturbed again. The positions of the continuous zero elements of the disturbed initial watermark are replaced with the binary representation of the position information of the two privacy parts to form the final watermark matrix W. The corresponding watermark embedding matrix B is generated according to this watermark matrix, and then the watermark embedding matrix is ​​multiplied with the watermark matrix to obtain the watermark part BW. In this step, the corresponding watermark embedding matrix B1 is also generated according to the initial watermark w1 of the level I privacy position. This matrix is ​​used as the key of the level I privacy position for subsequent decoding operations.

[0059] The watermark part BW is added to the compressed sensing encrypted data to generate ciphertext, and then the ciphertext is further encrypted using the disturbance matrix Add to prevent malicious attacks, and the operation on the encoding end is completed.

[0060] At the decoding end, the administrator user needs to use key B1 to restore the watermark part of the signal for the recovery of the level I privacy area, restore the watermark of the corresponding size, and treat the rest as noise. The restored images of the level I privacy area and the non-privacy area are obtained; the advanced authorized user mainly uses key B to restore the disturbed watermark information for the recovery of level II privacy, and then performs related operations to obtain the real watermark information, and obtains the corresponding mask and the position of the corresponding privacy area according to the watermark information, so as to obtain the entire image data.

[0061] After completing compressed sensing encryption and watermark embedding, the security of image privacy protection is further improved through a perturbation operation. The perturbation generation process first sets the sampling interval and the initial parameters of the chaotic map to ensure the diversity and unpredictability of the chaotic sequence. Specifically, the system calculates the total length of the perturbation sequence according to the width and height of the image and initializes three interrelated chaotic variables. Through a series of nonlinear iterative formulas, these chaotic variables generate highly random and complex sequences, which enhances the chaotic characteristics of the system. Subsequently, one of the main chaotic sequences is selected and downsampled according to the set sampling interval to obtain a compact and random chaotic vector. This downsampled sequence is reconstructed into a chaotic matrix Add that matches the image size. Finally, the generated chaotic matrix Add is matrix multiplied with the ciphertext Yw after watermark embedding to obtain the final perturbed ciphertext Yec. This perturbation process further confuses the ciphertext data structure by introducing additional randomness and complexity, greatly improving the security and anti-attack capability of the encrypted image.

[0062] The perturbation generation step uses nonlinear iteration to generate highly random chaotic sequences by setting reasonable sampling intervals and initial parameters. After downsampling and reconstruction, these sequences form a chaotic matrix that matches the image size. The chaotic matrix is ​​multiplied with the ciphertext after watermark embedding to achieve further randomization of the ciphertext. This process not only increases the complexity and unpredictability of the ciphertext, but also effectively resists various statistical analysis and cracking attacks, ensuring the high security of image privacy.

[0063] Example 2 In one embodiment of the present disclosure, a multi-position overlapping image privacy protection system based on compressed sensing is provided, comprising: A data acquisition module, used to acquire the original image, pre-process the original image, and divide the privacy part and the non-privacy part of different parts of the original image; An image reconstruction module is used to locate the overlapping private parts, generate an overlapping privacy mask for the overlapping private parts, generate a privacy mask for the non-overlapping private parts, use the generated overlapping privacy mask and privacy mask as a confusion matrix to confuse all the private parts, and use the confused private parts and the original non-private parts to reconstruct the image to obtain a reconstructed image; A watermark embedding module is used to perform compressed sensing sampling encryption on the reconstructed image to generate compressed sensing ciphertext, generate corresponding watermark matrices and watermark embedding matrices for different privacy parts according to the confusion matrix, embed the watermark embedding matrix and the watermark matrix into the compressed sensing ciphertext, and realize encryption of the compressed sensing ciphertext; The decoding and reconstruction module is used to decode and reconstruct different parts of the encrypted compressed sensing ciphertext using users with different levels of keys, thereby realizing the privacy protection of overlapping images in multiple locations.

[0064] Example 3 In one embodiment of the present disclosure, a computer program product is provided, including a computer program, and when the computer program is executed by a processor, the multi-position overlapping image privacy protection method based on compressed sensing is implemented.

[0065] Example 4 In one embodiment of the present disclosure, a non-transitory computer-readable storage medium is provided, wherein the non-transitory computer-readable storage medium is used to store computer instructions. When the computer instructions are executed by a processor, the multi-location overlapping image privacy protection method based on compressed sensing is implemented.

[0066] Example 5 In one embodiment of the present disclosure, an electronic device is provided, including: a processor, a memory, and a computer program; wherein the processor is connected to the memory, the computer program is stored in the memory, and when the electronic device is running, the processor executes the computer program stored in the memory, so that the electronic device executes the multi-location overlapping image privacy protection method based on compressed sensing.

[0067] The present disclosure is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present disclosure. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of the processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1A device that provides the functions specified in a block or multiple blocks.

[0068] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0069] Although the above describes the specific implementation methods of the present disclosure in conjunction with the accompanying drawings, it is not intended to limit the scope of protection of the present disclosure. Technical personnel in the relevant field should understand that on the basis of the technical solution of the present disclosure, various modifications or variations that can be made by those skilled in the art without creative work are still within the scope of protection of the present disclosure.

Claims

1. A privacy protection method for multi-position overlapping images based on compressed sensing, characterized in that: include: Acquire the original image, pre-process the original image, and divide the privacy part and the non-privacy part of different parts of the original image; Locate the overlapping private parts, generate overlapping privacy masks for the overlapping private parts, generate privacy masks for the non-overlapping private parts, use the generated overlapping privacy masks and privacy masks as confusion matrices to confuse all private parts, and reconstruct the image using the confused private parts and the original non-private parts to obtain a reconstructed image; Performing compressed sensing sampling encryption on the reconstructed image to generate compressed sensing ciphertext, generating corresponding watermark matrices and watermark embedding matrices for different privacy parts according to the confusion matrix, embedding the watermark embedding matrix and the watermark matrix into the compressed sensing ciphertext, and realizing encryption of the compressed sensing ciphertext; Users with different levels of keys can decode and reconstruct different parts of the encrypted compressed sensing ciphertext to achieve privacy protection of overlapping images in multiple locations.

2. The privacy protection method for multi-position overlapping images based on compressed sensing as claimed in claim 1, characterized in that: The original image is obtained and preprocessed to divide the original image into privacy parts and non-privacy parts of different parts, and represented by different matrices, wherein the privacy positions of the matrix of the non-privacy part are all zero, and the matrix of the privacy part is all zero except the privacy position.

3. The privacy protection method for multi-position overlapping images based on compressed sensing as claimed in claim 1, characterized in that: Get the location information of multiple private parts, record the location coordinates of the upper left corner and the lower right corner of the private parts, convert them into binary, perform corresponding binary processing on the converted binary values, so that the final value is 1 or -1, and the number of generated element bits is p; generate a matrix of the same size as the original image, and the elements in the matrix are randomly distributed -1, 1 values; multiply the masks corresponding to multiple private positions by the generated randomly distributed -1, 1 matrix respectively, and obtain the corresponding confusion matrix. At this time, only the location information of the corresponding private parts in the confusion matrix is ​​-1 or 1, and the rest of the positions are all zero.

4. The privacy protection method for multi-position overlapping images based on compressed sensing as claimed in claim 3, characterized in that: When processing the private part of a single position, a corresponding binary mask is generated as a privacy mask according to the respectively marked private parts; for the overlapping private parts, a new mask is separately marked to represent the overlapping parts, and the mask of the overlapping parts corresponding to each privacy area is replaced with the new mask as the overlapping privacy mask.

5. The privacy protection method for multi-position overlapping images based on compressed sensing as claimed in claim 1, characterized in that: The generated overlapping privacy masks and privacy masks are used as confusion matrices to confuse all private parts to obtain the information of the confused private positions. The information of the confused private parts is then added to the information of the non-sensitive parts to obtain the reconstructed image. The entire reconstructed image is then sampled and encrypted using compressed sensing.

6. The privacy protection method for multi-position overlapping images based on compressed sensing as claimed in claim 1, characterized in that: Generate corresponding watermark matrix and watermark embedding matrix for different privacy parts according to the confusion matrix, and embed the watermark embedding matrix and watermark matrix into compressed sensing ciphertext, including: The non-zero rows in the generated overlapping privacy mask and privacy mask matrix are removed to obtain two new matrices. All masks except the position of privacy area I in the overlapping privacy mask corresponding to the overlapping privacy area II are inverted. The transformed matrix is ​​called the initial watermark matrix, and the positions of consecutive zero elements in the initial watermark matrix are replaced with processed elements marked with binary codes with partial information of different privacy areas to form a watermark matrix. The corresponding watermark embedding matrix and left-zeroing matrix are generated according to the size of the watermark matrix. The watermark embedding matrix of the overlapping privacy part II is multiplied with the watermark matrix, and the obtained data is added to the compressed sensing ciphertext of compressed sensing to generate the final encrypted signal, and then the disturbance matrix is ​​added to the encrypted signal.

7. A multi-position overlapping image privacy protection system based on compressed sensing, characterized in that: include: A data acquisition module, used to acquire the original image, pre-process the original image, and divide the privacy part and the non-privacy part of different parts of the original image; An image reconstruction module is used to locate the overlapping private parts, generate an overlapping privacy mask for the overlapping private parts, generate a privacy mask for the non-overlapping private parts, use the generated overlapping privacy mask and privacy mask as a confusion matrix to confuse all the private parts, and use the confused private parts and the original non-private parts to reconstruct the image to obtain a reconstructed image; A watermark embedding module is used to perform compressed sensing sampling encryption on the reconstructed image to generate compressed sensing ciphertext, generate corresponding watermark matrices and watermark embedding matrices for different privacy parts according to the confusion matrix, embed the watermark embedding matrix and the watermark matrix into the compressed sensing ciphertext, and realize encryption of the compressed sensing ciphertext; The decoding and reconstruction module is used to decode and reconstruct different parts of the encrypted compressed sensing ciphertext using users with different levels of keys, thereby realizing privacy protection of overlapping images in multiple locations.

8. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the multi-position overlapping image privacy protection method based on compressed sensing described in any one of claims 1 to 6 is implemented.

9. A non-transitory computer-readable storage medium, characterized in that: The non-transitory computer-readable storage medium is used to store computer instructions. When the computer instructions are executed by the processor, the multi-position overlapping image privacy protection method based on compressed sensing as described in any one of claims 1-6 is implemented.

10. An electronic device, characterized in that: include: A processor, a memory and a computer program; wherein the processor is connected to the memory, the computer program is stored in the memory, and when the electronic device is running, the processor executes the computer program stored in the memory so that the electronic device implements the multi-position overlapping image privacy protection method based on compressed sensing as described in any one of claims 1-6.

Images