JPEG (Joint Photographic Experts Group) big data reversible data hiding method and system based on multi-dimensional features and differential firefly algorithm
Through the method based on multi-dimensional features and differential firefly algorithm, the balance of embedding capacity, image distortion and file extension in JPEG image reversible data hiding technology is solved, and efficient data hiding and image quality protection is achieved.
Patent Information
- Application Number
- CN202411598961.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-11-11
- Publication Date
- 2025-02-14
AI Technical Summary
In the reversible data hiding technology of JPEG images, how to balance the embedding capacity, image distortion and file expansion without destroying the file format has become a key issue.
Using a method based on multidimensional feature and differential firefly algorithm, the most similar DCT blocks were selected by constructing a multidimensional feature similarity evaluation model, and the most similar DCT blocks were selected for the construction of AC coefficient distribution model, and the distortion was preferred through Laplace distribution fitting. Smaller coefficients are used to embed data, and data hiding is finally achieved through optimal two-dimensional histogram mapping.
This method can improve the visual quality of the carrier image while keeping the JPEG file format unchanged, reduce file expansion, and achieve good embedding capacity and image quality balance.
Smart Images
Figure CN119442325A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of big data multimedia information security technology, and in particular to a JPEG big data reversible data hiding method and system based on multidimensional features and differential firefly algorithm. Background Art
[0002] With the rapid development of computer technology and the Internet, the number of digital images has grown exponentially, becoming an important information carrier in the process of big data informatization, and is widely used in social networks, electronic devices and other scenarios. With the widespread dissemination of images on various platforms, how to ensure the security, integrity and copyright protection of images in the storage, distribution and informatization processing environment of big data has become a key security requirement in the informatization environment. Therefore, reversible data hiding technology came into being. This technology not only supports the extraction of embedded data, but also can restore the original content of the image without loss, meeting highly sensitive applications such as military communications, legal evidence collection and medical imaging that have strict requirements on data integrity.
[0003] As a widely used compressed image format, JPEG has become the mainstream format for image storage and transmission in social media and electronic devices in the big data environment due to its high compression ratio and good visual quality. However, the lossy compression characteristics of JPEG pose significant challenges to reversible data hiding. During the JPEG compression process, the image is converted to a frequency domain representation through discrete cosine transform (DCT), and redundant information is removed to reduce the file size. This process greatly limits the modifiable space of DCT coefficients. Directly modifying these coefficients not only easily leads to a significant expansion of file size, increasing storage and transmission costs in big data scenarios, but may also introduce visual distortion and affect image quality. In addition, the block processing and complex coding structure of JPEG images require special caution when embedding data to avoid destroying the file format, resulting in decoding errors or unexpected visual effects.
[0004] Therefore, how to achieve a good balance between embedding capacity, image distortion and file expansion without destroying the file format has become a key issue in the research of reversible data hiding technology for JPEG images. Summary of the invention
[0005] The present invention provides a JPEG big data reversible data hiding method and system based on multidimensional features and differential firefly algorithm, so as to improve the visual quality of carrier images and reduce file expansion while keeping the JPEG file format unchanged.
[0006] In order to achieve the above object, the present invention adopts the following technical scheme.
[0007] The present invention provides a JPEG big data reversible data hiding method and system based on multidimensional features and differential firefly algorithm, comprising:
[0008] S1 combines the texture, coefficient distribution and spatial characteristics of the DCT coefficient block to build a multi-dimensional feature similarity evaluation model;
[0009] S2 determines the most suitable number k of DCT blocks involved in constructing the AC coefficient distribution model through the differential firefly optimization algorithm;
[0010] S3 uses a multi-dimensional feature similarity evaluation model to select the most similar k DCT blocks for the construction of the AC coefficient distribution model;
[0011] S4 estimates the distortion value of each AC coefficient by fitting the Laplace distribution, and preferentially selects the coefficient with smaller distortion for data embedding;
[0012] S5 constructs a two-dimensional histogram and selects an optimal two-dimensional histogram mapping according to the statistical characteristics of the AC coefficient pairs;
[0013] S6 embeds the secret data into the carrier image according to the optimal two-dimensional histogram mapping relationship, and finally obtains a marked JPEG image.
[0014] Furthermore, the step S1 specifically includes the following steps:
[0015] Divide the original JPEG image of size M×N into K 8×8 sub-blocks;
[0016] The DCT coefficients are extracted and scanned in zigzag order to form an AC coefficient matrix for embedding data;
[0017] The texture similarity is calculated by summing the quantization steps of all non-zero AC coefficients in the DCT block;
[0018] The coefficient distribution similarity is constructed by combining the similarity of the number of low, medium and high frequency non-zero AC coefficients, the similarity of the position of the last non-zero AC coefficients, and the similarity of the block template;
[0019] The Euclidean distance is used to measure the spatial distribution similarity of DCT blocks in the image;
[0020] Furthermore, a similarity evaluation model is constructed based on multi-dimensional features such as texture, coefficient distribution and spatial similarity of the DCT coefficient block. The multi-dimensional feature similarity evaluation model is calculated using the following formula:
[0021]
[0022] Among them, D i,j Indicates the comprehensive similarity of two DCT blocks through texture similarity C i,j , coefficient distribution similarity Z i,j and position similarity Y i,jγ is a balance parameter used to adjust the weights of each similarity item.
[0023] Furthermore, in step S2, the optimal number k of DCT blocks for constructing the AC coefficient distribution model is determined by using a differential firefly optimization algorithm.
[0024] Initialize the firefly swarm, set the number of fireflies, maximum number of iterations, step size, and the influence factor between fireflies;
[0025] Each firefly simulates the embedding process according to its current number of DCT blocks k, and calculates the peak signal-to-noise ratio (PSNR) as the fitness value;
[0026] According to the fitness value (PSNR) of each firefly, its luciferin concentration is updated. Fireflies with higher fitness values have higher luciferin concentrations, making them more attractive to other fireflies in the next step;
[0027] In each iteration, according to the difference in fluorescein concentration, the fireflies move to the fireflies with higher fluorescein concentration. By dynamically adjusting the step size step_size, the fireflies gradually concentrate in the global optimal area;
[0028] After several iterations, the firefly with the highest fitness value is finally selected, and its corresponding block number k is used as the optimal DCT block number.
[0029] Furthermore, the step S3 screens all DCT blocks of the JPEG image through a multi-dimensional feature similarity evaluation model to find k DCT blocks that are most similar in features for constructing an AC coefficient distribution model.
[0030] Furthermore, the step S4 uses Laplace distribution to perform fitting estimation on each AC coefficient, so as to calculate the distortion value of each coefficient.
[0031] Based on the k similar DCT block set selected by S3, the AC coefficients at each frequency are distributed and fitted. The following are the detailed steps:
[0032] By calculating the probability distribution function of the AC coefficients in each DCT block, according to the formula:
[0033]
[0034] Among them, λ u,y is the scale parameter associated with the frequency (u,v), and sign(x) is the sign function used to represent the sign of x.
[0035] Estimate the scale parameter associated with the frequency position (u,v) to fit the Laplace distribution;
[0036] The distortion value of each AC coefficient is estimated through the Laplace distribution model, and the coefficient with smaller distortion is used as the embedding candidate. The distortion value can be estimated by the following formula:
[0037]
[0038] Among them, λ i u,y It is the scale parameter at the frequency position (u, v), reflecting the distribution characteristics of the AC coefficients, and q(u, v) is the quantization step size at this position.
[0039] The AC coefficients with less distortion are preferentially selected for data embedding, thereby ensuring that the impact on image quality is minimal while embedding data.
[0040] Furthermore, the step S5 constructs a two-dimensional histogram and selects the best mapping mode from four valid two-dimensional histogram mappings by comparative analysis of the embedding capacity (EC) and the embedding distortion (ED).
[0041] According to the candidate AC coefficients, the AC coefficients are paired and a two-dimensional histogram is constructed;
[0042] The AC coefficient pairs are classified into five types: 1. (±1, ±1) coefficient pairs; 2. (±1, ±2) coefficient pairs; 3. (±2, ±1) coefficient pairs; 4. (±2, ±2) coefficient pairs; 5. coefficient pairs of other values, and the frequency of each type of coefficient pairs is counted;
[0043] Based on different two-dimensional histogram mapping relationships, the embedding capacity (EC) corresponding to each type of AC coefficient pair is calculated:
[0044]
[0045] Where H(u) is the frequency of pixel pairs in the two-dimensional histogram, and f m (u) is the number of candidate outputs for the mapped pixel pair.
[0046] Based on different two-dimensional histogram mapping relationships, the embedding distortion (ED) corresponding to each type of AC coefficient pair is calculated:
[0047]
[0048] in, It is the difference between the mapped pixel pair and the original pixel pair, reflecting the distortion caused by the embedding operation.
[0049] According to the ratio of embedding capacity (EC) and embedding distortion (ED), the optimal two-dimensional histogram mapping method is selected. The optimal mapping should make the embedding capacity as large as possible and the embedding distortion as small as possible.
[0050] Furthermore, the step S6 adjusts the AC coefficient pairs in each DCT block according to the optimal two-dimensional histogram mapping method selected in S5 to embed the secret signal data.
[0051] The updated DCT coefficient matrix is applied to the JPEG image and re-encoded to generate a marked JPEG image containing embedded information. The marked image contains the secret data but visually maintains the quality of the original image. BRIEF DESCRIPTION OF THE DRAWINGS
[0052] Figure 1 A JPEG big data reversible data hiding method and system flow chart based on multi-dimensional features and differential firefly algorithm in an embodiment;
[0053] Figure 2 It is an overall block diagram of the embedding process in the embodiment;
[0054] Figure 3 It is a flow chart of the differential firefly optimization algorithm in the embodiment;
[0055] FIG4 is an image of a lake and a boat used in a simulation experiment in an embodiment;
[0056] FIG5 is a Lake and Boat encrypted image used in the simulation experiment in the embodiment;
[0057] Figure 6 The visual quality changes of Lake and Boat images under different embedding capacities when the compression quality factor is 80 in the embodiment;
[0058] Figure 7 is the file expansion change under different embedding capacities when the compression quality factor of Lake and Boat images in the embodiment is 80. is the difference between the mapped pixel pair and the original pixel pair, reflecting the distortion caused by the embedding operation. DETAILED DESCRIPTION
[0059] The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, but should not be construed as limiting the present invention.
[0060] Figure 1 This is a reversible data hiding method and system flow chart for JPEG big data based on multidimensional features and differential firefly algorithm. Figure 2 This is the overall block diagram of the embedding process of this embodiment, refer to Figure 1 and Figure 2 , the method comprises the following steps:
[0061] S1 combines the texture, coefficient distribution and spatial characteristics of discrete cosine transform (DCT) coefficient blocks to build a multi-dimensional feature similarity evaluation model.
[0062] The original JPEG image is parsed to obtain the quantized DCT coefficient block. Each 8×8 DCT block contains 64 DCT coefficients, including 1 DC coefficient and 63 AC coefficients.
[0063] A multi-dimensional feature similarity evaluation model is constructed to calculate the texture similarity, coefficient distribution similarity and spatial similarity of DCT sub-blocks. The specific steps are as follows.
[0064] For each DCT block, its texture similarity is calculated, and the texture similarity is represented by the sum of the quantization steps of all non-zero AC coefficients in the DCT block.
[0065] The AC coefficients of each DCT block are divided into low-frequency, medium-frequency and high-frequency regions according to their frequencies. The number of non-zero AC coefficients in each region is counted respectively. The coefficient distribution similarity model is constructed by combining the position of the last non-zero AC coefficient in the block and the template similarity. The coefficient distribution similarity is calculated by the following formula:
[0066]
[0067] Among them, C i,j Indicates the similarity of coefficient distribution between two DCT blocks, including template matching similarity T i,j , similarity of number of non-zero AC coefficients N i,j and the last non-zero AC coefficient position similarity P i,j .
[0068] For each DCT block, the spatial similarity is calculated by calculating the Euclidean distance between DCT blocks.
[0069] Combining the above texture similarity, coefficient distribution similarity (including the number of low-frequency, medium-frequency, and high-frequency AC coefficients, template matching similarity, and the position of the last non-zero AC coefficient) and spatial characteristics, a multi-dimensional feature similarity model is constructed. This model is used to screen the most similar DCT blocks in subsequent steps. The final similarity evaluation formula is as follows:
[0070]
[0071] Among them, D i,j Indicates the comprehensive similarity of two DCT blocks through texture similarity C i,j , coefficient distribution similarity Z i,j and position similarity Y i,j γ is a balance parameter used to adjust the weights of each similarity item.
[0072] S2, determine the most suitable number of DCT blocks k for constructing the AC coefficient distribution model through the differential firefly optimization algorithm, refer to Figure 3 shown.
[0073] Initialize the relevant parameters of the differential firefly optimization algorithm, such as the number of fireflies, fluorescein concentration volatility factor, step size, and field change rate.
[0074] The DCT block selection scheme of the initial fireflies is generated randomly, and each firefly randomly selects a block number as its initial position. Each firefly embeds data according to the currently selected block number at each iteration, and calculates the peak signal-to-noise ratio (PSNR) after the simulated embedding, which is used as the fitness value. The higher the fitness value, the smaller the image distortion of the corresponding DCT block number selection scheme when embedding data.
[0075] In each iteration, the firefly compares with its neighboring fireflies, selects the firefly with a higher fluorescein concentration as a reference, and dynamically adjusts the position of the current firefly based on its position. The fluorescein concentration is updated according to the fitness value of each firefly to ensure that the firefly with a higher fluorescein concentration can attract the neighboring fireflies.
[0076] After multiple iterations, the algorithm selects the number of blocks corresponding to the fireflies with the highest PSNR value as the optimal solution, and determines this number k as the optimal number of DCT blocks for constructing the AC coefficient distribution model.
[0077] S3 uses a multi-dimensional feature similarity evaluation model to select the most similar k DCT blocks for constructing the AC coefficient distribution model.
[0078] Based on the multi-dimensional feature similarity evaluation model, the similarity between each DCT block and all other DCT blocks is calculated.
[0079] The k DCT blocks that are most similar to other blocks are selected as the basis for the subsequent AC coefficient distribution model.
[0080] By fitting the Laplace distribution of the AC coefficients in the selected k most similar DCT blocks, the distortion value of each AC coefficient is estimated, and the coefficients with smaller distortion are preferentially selected for data embedding according to the size of the distortion value. This process ensures that the image distortion during the embedding process is minimized while maintaining a high embedding capacity.
[0081] For the AC coefficients in the k DCT blocks selected from step S3, the scale parameters related to the frequency position (u, v) are estimated by calculating the probability distribution function of the AC coefficients in each DCT block, so as to fit the Laplace distribution according to the formula:
[0082]
[0083] Among them, λ u,y is the proportional parameter related to the frequency (u,v), and sign(x) is the sign function used to indicate the sign of x.
[0084] The distortion value of each AC coefficient is estimated through the Laplace distribution model, and the coefficient with smaller distortion is used as the embedding candidate. The distortion value can be estimated by the following formula:
[0085]
[0086] Among them, λ i u,y It is the scale parameter at the frequency position (u, v), reflecting the distribution characteristics of the AC coefficients, and q(u, v) is the quantization step size at this position.
[0087] After estimating the distortion of each AC coefficient, the coefficients are sorted according to the size of the distortion value. According to the sorting result, the AC coefficient with smaller distortion is preferentially selected as the target of data embedding. In this way, the impact on the image during the data embedding process can be minimized and high image quality can be maintained.
[0088] S5: construct a two-dimensional histogram, and select the optimal two-dimensional histogram mapping according to the statistical characteristics of the AC coefficient pairs.
[0089] Since the AC coefficients with smaller distortion have been preferentially selected in step S4, these coefficients are paired to form a two-dimensional histogram.
[0090] According to the paired AC coefficients, the occurrence frequency of each pair of AC coefficients in all DCT blocks is counted.
[0091] To improve computational efficiency, we classify AC coefficient pairs according to their absolute values.
[0092] Category 1: coefficient pairs (±1,±1) with an absolute value of 1;
[0093] Category 2: coefficient pairs (±1, ±2) with an absolute value of 2;
[0094] Category 3: coefficient pairs (±2,±1) with an absolute value of 3;
[0095] Category 4: coefficient pairs (±2,±2) with an absolute value of 3;
[0096] Category 5: Other AC coefficient pairs with larger absolute values.
[0097] For the four valid two-bit histogram mappings, we need to determine the most suitable mapping scheme. The mapping goal is to find a way to maximize the embedding capacity (EC) and minimize the embedding distortion (ED). The specific mapping selection steps are as follows:
[0098]
[0099] Where H(u) is the frequency of pixel pairs in the two-dimensional histogram, and f m (u) is the number of candidate outputs for the mapped pixel pair.
[0100]
[0101] in, It is the difference between the mapped pixel pair and the original pixel pair, reflecting the distortion caused by the embedding operation.
[0102] The embedding capacity (EC) and embedding distortion (ED) under each mapping are calculated one by one, and finally the mapping scheme with the smallest ED / EC ratio is selected as the optimal two-dimensional histogram mapping.
[0103] S6. According to the optimal two-dimensional histogram mapping relationship, the secret data is embedded into the carrier image, and a marked JPEG image is output.
[0104] The DCT coefficients are modified according to the optimal two-dimensional histogram mapping.
[0105] After all AC coefficient pairs are adjusted, the adjusted DCT coefficients are encoded to generate the final JPEG image file, which contains the secret data while maintaining the visual quality of the original image while ensuring the file size.
[0106] In order to verify the actual effect of the solution of the present invention, this embodiment verifies and illustrates the solution through performance analysis and simulation experiments.
[0107] This embodiment is a simulation example of a JPEG big data reversible data hiding method and system based on multidimensional features and differential firefly algorithm. The experimental simulation uses 512×512 grayscale images of Lake and Boat as test images, and uses the JPEG big data reversible data hiding method and system based on multidimensional features and differential firefly algorithm of the present invention to obtain the following results:
[0108] The calculation formula of PSNR value is as follows:
[0109]
[0110] Among them, MSE stands for mean square error, which is used to measure the difference between the image before and after embedding, and MAX represents the maximum possible value of the image pixel.
[0111] Figures 4 and 5 are the original and secret images of the Lake and Boat test images, respectively. Embedding secret data does not change the visual effects and characteristics of the images.
[0112] Figure 6 The peak signal-to-noise ratio (PSNR)-embedding capacity result diagram of the encrypted image obtained by the method of the present invention when the compression quality factor of Lake and Boat images is 80. It can be seen from the data in the figure that with the increase of embedding capacity, the PSNR value gradually decreases, indicating that the larger the amount of embedded data, the greater the impact on image quality. However, even at a high embedding capacity, the method of the present invention can still maintain a high PSNR value, indicating that it has little loss on image quality and has good visual effects and embedding performance.
[0113] Figure 7 This is a file expansion-embedding capacity result diagram of the encrypted image obtained by the method of the present invention when the compression quality factor of Lake and Boat images is 80. From the data in the figure, it can be seen that the method of the present invention has excellent file size management capabilities while meeting the embedding requirements.
[0114] In summary, the present invention proposes a JPEG big data reversible data hiding method and system based on a multidimensional feature similarity evaluation model and a differential firefly optimization algorithm. First, a multidimensional feature similarity model combining DCT block texture, coefficient distribution and spatial characteristics is constructed to measure the similarity between DCT blocks. Further, the differential firefly optimization algorithm is used to determine the number of DCT blocks k that are most suitable for participating in the construction of the AC coefficient distribution model, and the most similar k blocks are screened out for Laplace distribution fitting, and the distortion value of each AC coefficient is estimated, and the coefficient with smaller distortion is preferentially selected to ensure image quality. In addition, by selecting the optimal two-dimensional histogram mapping for data embedding, an embedding effect with low loss and low file expansion is achieved. This scheme takes into account the embedding capacity, image quality and file extension size, ensures the efficiency and practicality of data hiding, and effectively meets the security and integrity requirements in the big data information environment.
Claims
1. A JPEG big data reversible data hiding method and system based on multidimensional features and differential firefly algorithm, characterized in that: The method comprises: S1. Combining the texture, coefficient distribution and spatial characteristics of discrete cosine transform (DCT) coefficient blocks, a multi-dimensional feature similarity evaluation model is constructed; S2, determining the most suitable number k of DCT blocks for constructing the AC coefficient distribution model by using a differential firefly optimization algorithm; S3, using a multi-dimensional feature similarity evaluation model to select the most similar k DCT blocks for constructing an AC coefficient distribution model; S4, estimating the distortion value of each AC coefficient through Laplace distribution fitting, and preferentially selecting coefficients with smaller distortion for data embedding; S5, constructing a two-dimensional histogram, and selecting an optimal two-dimensional histogram mapping according to the statistical characteristics of the AC coefficient pairs; S6. According to the optimal two-dimensional histogram mapping relationship, the secret data is embedded into the carrier image, and a marked JPEG image is output.
2. According to claim 1, a JPEG big data reversible data hiding method and system based on multidimensional features and differential firefly algorithm is characterized in that: The step S1 specifically includes: S11, parsing the original JPEG image to obtain a quantized DCT coefficient block; S12. Constructing a multi-dimensional feature similarity evaluation model based on the multiple DCT coefficient blocks.
3. The method and system for reversible data hiding of JPEG big data based on multidimensional features and differential firefly algorithm according to claim 2 is characterized in that: The step S1 constructs a multi-dimensional feature similarity evaluation model according to the texture, coefficient distribution and spatial characteristics of the DCT coefficient block: Texture similarity: reflects the difference in details and complexity in the image block by calculating the sum of the quantization steps of all non-zero AC coefficients in the DCT block; Spatial similarity: By calculating the spatial position difference of DCT blocks in the image, the Euclidean distance is used to measure the spatial position difference between DCT blocks to describe the relative distribution of DCT blocks in space; Coefficient distribution similarity: The difference in the number of non-zero AC coefficients of low-frequency, medium-frequency, and high-frequency AC coefficients reflects the difference in the amount of information of each DCT block in different frequency bands; the position of the last non-zero AC coefficient measures the characteristics of information distribution in the DCT block; template matching difference, counts the number of non-zero AC coefficients of each DCT block at the same position with other blocks to measure the structural similarity between blocks. Coefficient distribution similarity is calculated using the following formula: ， Among them, C i,j Indicates the similarity of coefficient distribution between two DCT blocks, including template matching similarity T i,j , similarity of number of non-zero AC coefficients N i,j and the last non-zero AC coefficient position similarity P i,j .
4. The method and system for reversible data hiding of JPEG big data based on multidimensional features and differential firefly algorithm according to claim 3 is characterized in that: The similarity evaluation model in step S1 is calculated by the following formula: ， Among them, D i,j Indicates the comprehensive similarity of two DCT blocks through texture similarity C i,j , coefficient distribution similarity Z i,j and position similarity Y i,j γ is a balance parameter used to adjust the weights of each similarity item.
5. The method and system for reversible data hiding of JPEG big data based on multidimensional features and differential firefly algorithm according to claim 4 is characterized in that: The differential firefly optimization algorithm in step S2 comprises the following steps: S21, initialize the firefly swarm, set the number of fireflies, maximum number of iterations, fluorescein concentration, step size and neighborhood change rate; S22, each firefly randomly selects an initial block number within a predefined block number range; S23, calculating the fitness value of each firefly by simulating the embedding process, where the fitness value is based on the effect of the number of blocks on the peak signal-to-noise ratio (PSNR) of the image; S24, updating the fluorescein concentration of each firefly according to its fitness value, and adjusting the position of the fireflies by comparing the fluorescein concentrations, and choosing to move closer to the fireflies with higher fitness values; S25. In each iteration process, dynamically adjust the decision radius of the firefly to control its position change; S26. After several iterations, the number of blocks corresponding to the firefly with the highest fitness value is selected as the optimal value k.
6. The method and system for reversible data hiding of JPEG big data based on multidimensional features and differential firefly algorithm according to claim 5, characterized in that: In step S3, the DCT coefficient blocks involved in fitting are determined by using the similarity evaluation model and the number k of the best DCT blocks.
7. The JPEG big data reversible data hiding method and system based on multidimensional features and differential firefly algorithm according to claim 6 is characterized in that: The distortion value in step S4 is estimated by fitting the Laplace distribution. The probability distribution function of the AC coefficients in each DCT block is calculated to estimate the scale parameter associated with the frequency position (u, v), thereby fitting the Laplace distribution according to the formula: ， Among them, λ u,y is the scale parameter associated with the frequency (u,v), and sign(x) is the sign function used to represent the sign of x.
8. The JPEG big data reversible data hiding method and system based on multidimensional features and differential firefly algorithm according to claim 7, characterized in that: In step S4, based on the fitting result of the Laplace distribution, AC coefficients with smaller distortion values are preferentially selected to embed the secret data, so as to minimize the impact on the image quality during the embedding process.
9. The JPEG big data reversible data hiding method and system based on multidimensional features and differential firefly algorithm according to claim 8, characterized in that: In step S5, a two-dimensional histogram is constructed to classify the AC coefficients according to their statistical characteristics, which specifically includes: S81, pairing AC coefficients according to the AC coefficients with smaller distortion that are preferentially selected and constructing a two-dimensional histogram; S82, classify AC coefficient pairs into five types, namely:
1. (±1, ±1) coefficient pairs; 2. (±1, ±2) coefficient pairs; 3. (±2, ±1) coefficient pairs; 4. (±2, ±2) coefficient pairs; 5. coefficient pairs of other values; S83. Count the frequencies of various coefficient pairs as a basis for subsequent selection of the optimal two-dimensional histogram mapping, so as to ensure that during the embedding process, a two-dimensional histogram mapping relationship with less distortion and larger embedding capacity is selected to embed the secret data.
10. The method and system for reversible data hiding of JPEG big data based on multidimensional features and differential firefly algorithm according to claim 9, characterized in that: In step S5, a rate-distortion model is constructed to select the optimal two-dimensional histogram mapping according to the ratio of embedding capacity (EC) to embedding distortion (ED). The calculation formulas for embedding capacity and embedding distortion are: ， ， Where H(u) is the frequency of pixel pairs in the two-dimensional histogram, and f m (u) is the number of candidate outputs for the mapped pixel pairs, It is the difference between the mapped pixel pair and the original pixel pair, reflecting the distortion caused by the embedding operation.
11. The JPEG big data reversible data hiding method and system based on multidimensional features and differential firefly algorithm according to claim 10, characterized in that: In step S5, a rate-distortion model is constructed according to the embedding capacity and the embedding distortion to solve the following optimization problem: ， In order to ensure the embedding capacity EC(f m )≥g p Under the premise of m , where g p is the minimum required embedding capacity.
12. The method and system for reversible data hiding of JPEG big data based on multidimensional features and differential firefly algorithm according to claim 11, characterized in that: In step S6, the secret information is embedded into the AC coefficients of the carrier image using the optimal two-dimensional histogram mapping relationship, and a marked JPEG image is finally generated, ensuring that the visual quality and file size of the file during the embedding process are close to the original image.