An adaptive high-capacity steganographic method for noisy images

Abstract

The present invention discloses an adaptive high-capacity steganographic method for noise images, comprising: inputting an image and preprocessing the image; embedding data into the preprocessed image to obtain a secret image G; dividing the secret image G into s*s non-overlapping blocks, and decrypting the data and restoring the stego data. The present invention uses a genetic algorithm to adaptively select image embedding points to achieve high-quality steganography and ensure good invisibility of the stego image; at the same time, in order to increase the stego capacity, the BBE algorithm is used for compression to reduce the space occupied by the marker bit to increase the embedding capacity; when embedding, the algorithm embeds two bits of secret data at a time, and embeds the secret information according to the conversion rules known to both receiving parties, thereby ensuring the security of the data, and this method does not require the participation of the encryption key, reducing its occupied space.

Description

Technical Field

[0001] The present invention relates to the technical field of image processing, and in particular to an adaptive high-capacity steganographic method for noise images. Background Art

[0002] With the rapid development of data communication technology, the widespread use of various digital information processing technologies and the popularization of the Internet, various media information such as pictures, texts, sounds and images can be quickly and conveniently transmitted in various communication networks, which provides significant convenience for the compression, storage, copying, processing and utilization of information. At the same time, the copyright and content protection of digital media works have become increasingly prominent; on the other hand, the rapid development of communication networks has made the transmission and exchange of various types of information easier and more difficult to monitor, which has posed a great challenge to the protection of information in the military, commercial and civilian fields. As an information security technology, information hiding technology has been adopted in many fields. Among them, the use of information hiding technology for secure and covert communication has important application value in the military, intelligence, and government confidential departments. For example, using spread spectrum communication makes it difficult for the enemy to detect and interfere with communication signals. Information hiding technology can also be used to implement camouflage and deceive the enemy's communication reconnaissance system. Information hiding is to hide a meaningful information in another information called a public carrier (Cover) to obtain a hidden carrier S (Stego Cover). The illegal person does not know whether other information is hidden in this ordinary information, and even if he knows, it is difficult to extract or remove the hidden information. The carrier used can be text, image, sound, video, etc. In order to increase the difficulty of attack, encryption and information hiding technology can also be combined, that is, first encrypt the message M to obtain the ciphertext message M', and then hide M' in the carrier C to ensure the security and reliability of data transmission. The general algorithm can maintain good invisibility and anti-interference after realizing data steganography.

[0003] Existing steganographic algorithms can be roughly divided into two categories: spatial domain steganography and frequency domain steganography. Among them, spatial domain steganography is to achieve the purpose of information embedding by changing the pixel values ​​in the spatial domain of the image. Typical algorithms include the LSB steganography algorithm proposed by Walter Bender et al. and the MLSB (Multiple LSB) replacement steganography proposed by Charles Kurak et al. This type of algorithm replaces the lowest bit with secret information, and the image content information is easily lost, and the ability to resist attacks is poor. Spatial domain adaptive steganography is a type of steganographic algorithm with relatively high security. It usually determines which image data to change and how to change it based on some local (such as local texture complexity, etc.) or global statistical characteristics (such as joint probability distribution, etc.) of the image to minimize the embedding distortion. For example, Mandal et al. proposed a steganographic method combining adaptive LSB replacement and pixel-value differencing (PVD). The grayscale image is divided into 3×3 or 3×3 plus 2×2 pixel blocks. In a block, a pixel is called a reference pixel, which replaces the 4 least significant bits. Sahu et al. proposed a steganographic method based on pixel value difference and modulus function (Pixelvalue difference and modulus function). The image steganography method of using a pixel-by-pixel module function (PVDMF) is mainly used to improve the peak signal-to-noise ratio (PSNR) and increase the embedding capacity (embedding capacity, HC). Frequency domain steganography embeds information by changing the value of frequency domain coefficients, such as DCT (Discrete Cosine Transform) and DWT (Discrete Wavelet Transform). According to the selection rules of the embedding position and the coefficient change mode, frequency domain steganography can be roughly divided into replacement steganography, statistical characteristic preservation steganography and adaptive steganography. Thai et al. proposed the use of quantized discrete cosine transform (Discrete cosine transform) to embed information. transform (DCT) coefficients and used them to detect hidden information in JPEG images. By formulating hidden data detection as a hypothesis test, the most powerful likelihood ratio test for steganalysis of the Jsteg algorithm was studied, and its statistical performance was theoretically established. Luo et al. proposed two steganalysis methods to estimate the modification rate of F5 steganography and its improved version. These methods are usually used for confidentiality in hidden images. The estimated modification rate can be used to distinguish the hidden image marked by F5 steganography or its improved version from the original image, etc. Summary of the invention

[0004] The purpose of this section is to summarize some aspects of embodiments of the present invention and briefly introduce some preferred embodiments. Some simplifications or omissions may be made in this section and the specification abstract and the invention title of this application to avoid blurring the purpose of this section, the specification abstract and the invention title, and such simplifications or omissions cannot be used to limit the scope of the present invention.

[0005] In view of the above existing problems, the present invention is proposed.

[0006] Therefore, the technical problem solved by the present invention is that the prior art has low steganographic quality and capacity, and low data security.

[0007] To solve the above technical problems, the present invention provides the following technical solutions: input an image and preprocess the image; embed data into the preprocessed image to obtain a secret image G; divide the secret image G into s*s non-overlapping blocks, and decrypt the data and recover the steganographic data.

[0008] As a preferred solution of the adaptive high-capacity steganography method for noisy images described in the present invention, the preprocessing step includes filtering and denoising the area with high noise corresponding image frequency through the BM3D algorithm; comparing the data obtained after filtering by the BM3D algorithm with the optimal embedding threshold T obtained by the genetic algorithm, finding the points where the noise value is higher than the preset value and generating global candidate steganographic positions, and obtaining the marked bit plane B; replacing the second bit plane of the image pixel with the marked bit plane B in LSB mode to obtain the masked image I1.

[0009] As a preferred solution of the adaptive high-capacity steganographic method for noise images described in the present invention, the input image I is decomposed into bit planes to obtain P binary images, wherein the calculation formula for the corresponding bit of the pixel in the k-th bit plane is:

[0010]

[0011] Among them, i, j represent the corresponding position of the pixel, i-th row, j-th column, mod2 represents the remainder when divided by 2, I (k) (i, j) represents the pixel value of the i-th row and j-th column in the k-th bit plane of plane I.

[0012] As a preferred solution of the adaptive high-capacity steganographic method for noise images described in the present invention, wherein: embedding data into the preprocessed image to obtain the secret image G includes: performing bit plane decomposition on the mask image I1, reorganizing the secret information into a matrix sequence message, and extracting the second bit plane of the mask image I1, namely, the marking bit plane B; if B(i, j) = 0, no processing is performed at the position (i, j); otherwise, firstly judging whether the secret information m(1, j) and m(2, j) are equal, if they are equal, embedding data "1" into the same bit of the third bit plane, otherwise embedding data "0"; after the third bit plane is embedded with "1", detecting the first bit of secret data, when m(1, i) = 1, embedding "1" into the corresponding bit of the first bit plane, otherwise embedding "0" into the corresponding bit of the first bit plane when m(1, i) = 0; similarly, when the third bit plane is "0", it can be embedded into the first bit plane in the same way; finally, embedding the data of the first and third bit planes is completed to obtain the stego-image G1, and recording the embedding capacity N at the same time.

[0013] As a preferred solution of the adaptive high-capacity steganography method for noise images described in the present invention, it also includes: dividing the mask image I1 into s*s non-overlapping blocks, obtaining the block classification mark type and the embeddable capacity CB according to the BBE algorithm; recording the redundant part after compression, that is, the space that can be embedded again, and representing the position that can be embedded again in the block with "1", that is, F(i,j)=1, otherwise, the position that cannot be embedded again F(i,j)=0; recording the data E at the block classification mark position in the original block, superimposing it into the message as the secret information embedded in the second and fourth bit planes, and calculating the size length of E as k, expressed as message1=[E,message(1:2,N+1:end)]; repeating the above process until all blocks are compressed, obtaining the re-embeddable data mark bit plane F and the embeddable capacity C that can be increased b .

[0014] As a preferred solution of the adaptive high-capacity steganographic method for noise images described in the present invention, it also includes: embedding the second and fourth bit planes: dividing the stego image G1 after embedding the first and third bit planes into non-overlapping blocks of the same size s*s, and embedding them into the second and fourth bit planes according to the embedding process, recording the embedding capacity N1, and marking the bit plane from B to F; when all blocks are completely embedded with secret information, the secret image G is obtained, and the embedded capacity is recorded as Num=N+N1-k.

[0015] As a preferred solution of the adaptive high-capacity steganographic method for noise images described in the present invention, the secret image G is divided into s*s non-overlapping blocks, and data decryption and recovery of the stegographed data include: judging whether the first bit of data, i.e., m(1, k), is 1; then judging whether the second bit of data, m(2, k), is 1. When m(1, k)=1, if it is 1, it means that the embedded secret data is "11", otherwise it is "00"; and for m(1, k)=0, if m(2, k)=1, it means that the embedded data is "10", otherwise it is "01"; and so on, repeating the cycle to recover each column of the data M, and reorganizing to obtain the corresponding original data MB, and it can be known that MB is the combination of the data E and message1 mentioned in the above step of embedding.

[0016] As a preferred solution of the adaptive high-capacity steganography method for noise images described in the present invention, it also includes: according to the BBE algorithm, using the block identification bit data of the second bit plane to restore the second bit plane data; finding the mark bitmap data of the corresponding bit from the extracted data MB through the index of the non-embeddable block, and completely extracting the bit mark bitmap B embedded in the second bit plane b ; According to the extracted bit mark bitmap B b And the bit mark map B for determining the embedded position f , complete the extraction of data mess1 of the second and fourth planes of the secret image G.

[0017] As a preferred solution of the adaptive high-capacity steganographic method for noise images described in the present invention, it also includes: the bit mark map B b After the complete extraction, according to the embedding method of the first and third planes, the first and third plane information data mess2 are extracted according to the extraction method in the extraction process; the bit mark map B is judged b (i,j) is 1. If it is 1, it means that there is data embedded in this position. Otherwise, there is no data embedded. B b When (i, j) = 1, the bit data m1(i) corresponding to the third bit plane of the secret image G is extracted; and so on, the first and third bit plane data are repeatedly extracted to obtain the data mess2, and the final extracted information is message = [mess1, mess2].

[0018] The beneficial effects of the present invention are as follows: the present invention proposes an algorithm for embedding secret information in blocks by finding the best embedding threshold in the noise image after BM3D denoising by using a genetic algorithm; the method uses a genetic algorithm to realize adaptive selection of image embedding points to achieve high-quality steganography and ensure good invisibility of the steganographic image; at the same time, in order to improve the steganographic capacity, the BBE algorithm is used for compression to reduce the space occupied by the mark bit to improve the embedding capacity; when embedding, the algorithm embeds two bits of secret data at a time, and embeds the secret information according to the conversion rules known to both receiving parties, thereby ensuring the security of the data, and this method does not require the participation of an encryption key and reduces its occupied space. The present invention has good robustness, invisibility and security, and can be used in radar, remote sensing image data transmission, information steganography and other fields. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative labor. Among them:

[0020] Figure 1 A basic flow chart of an adaptive high-capacity steganographic method for noise images provided by one embodiment of the present invention;

[0021] Figure 2 A schematic diagram of a basic data embedding process of an adaptive high-capacity steganographic method for noise images provided by an embodiment of the present invention;

[0022] Figure 3 A schematic diagram of an encrypted data recovery process of an adaptive high-capacity steganographic method for noise images provided by an embodiment of the present invention;

[0023] Figure 4 A schematic diagram of the first and third plane data extraction process of an adaptive high-capacity steganographic method for noise images provided by one embodiment of the present invention;

[0024] Figure 5 A schematic diagram of visual quality of a comparison test image of an adaptive high-capacity steganography method for noise images provided by an embodiment of the present invention when fully embedded;

[0025] Figure 6 A schematic diagram of statistical invisibility analysis results of each image by an algorithm of an adaptive high-capacity steganography method for noise images provided by an embodiment of the present invention;

[0026] Figure 7 A schematic diagram of PSNR comparison of various steganographic images at different embedding rates of an adaptive high-capacity steganographic method for noise images provided by one embodiment of the present invention;

[0027] Figure 8 A schematic diagram of a secret image being attacked in an adaptive high-capacity steganographic method for noise images provided by an embodiment of the present invention;

[0028] Fig. 9 A schematic diagram of secret image extraction of an adaptive high-capacity steganography method for noisy images provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0029] In order to make the above-mentioned purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in the art without creative work should fall within the scope of protection of the present invention.

[0030] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0031] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive with other embodiments.

[0032] The present invention is described in detail with reference to schematic diagrams. When describing the embodiments of the present invention, for the sake of convenience, the cross-sectional diagrams showing the device structure will not be partially enlarged according to the general scale, and the schematic diagrams are only examples, which should not limit the scope of protection of the present invention. In addition, in actual production, the three-dimensional dimensions of length, width and depth should be included.

[0033] At the same time, in the description of the present invention, it should be noted that the directions or positional relationships indicated by the terms "upper, lower, inner and outer" are based on the directions or positional relationships shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific direction, be constructed and operated in a specific direction, and therefore cannot be understood as limiting the present invention. In addition, the terms "first, second or third" are only used for descriptive purposes and cannot be understood as indicating or implying relative importance.

[0034] In the present invention, unless otherwise clearly specified and limited, the terms "install, connect, connect" should be understood in a broad sense, for example: it can be a fixed connection, a detachable connection or an integral connection; it can also be a mechanical connection, an electrical connection or a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0035] Example 1

[0036] Reference Figures 1 to 4 , which is an embodiment of the present invention, provides an adaptive high-capacity steganographic method for noisy images, comprising:

[0037] S1: input image and preprocess the image;

[0038] It should be noted that in order to achieve better data steganography and find a suitable embedding position, the collected images need to be preprocessed to achieve data steganography; the preprocessing steps include:

[0039] The noise corresponds to the area with higher image frequency, and the BM3D algorithm is used to filter and denoise the area with higher noise corresponding image frequency;

[0040] The data obtained after filtering by the BM3D algorithm is compared with the optimal embedding threshold T obtained by the genetic algorithm, and the points where the noise value is higher than the preset value are found and the global candidate steganographic positions are generated to obtain the marked bit plane B;

[0041] The mark bit plane B replaces the second bit plane of the image pixel in LSB mode to obtain the mask image I1.

[0042] S2: embed the preprocessed image into a secret image G;

[0043] It should be noted that the input image I is decomposed into bit planes to obtain P binary images, where the calculation formula for the corresponding bit of the pixel in the kth bit plane is:

[0044]

[0045] Among them, i, j represent the corresponding position of the pixel, i-th row, j-th column, mod2 represents the remainder when divided by 2, I (k) (i, j) represents the pixel value of the i-th row and j-th column in the k-th bit plane of plane I.

[0046] After preprocessing the original image I, the mask image I1 is obtained, where the second bit plane of the mask image I1 contains the marker bit B information; find the embedding point position, and the data is Figure 2 The embedding process first embeds the pixels into the first and third planes of the image to complete the first step of data embedding.

[0047] Specifically, the cover image I1 is decomposed into bit planes, the secret information is reorganized into a matrix sequence message, and the second bit plane of the cover image I1, namely, the marking bit plane B, is extracted;

[0048] If B(i, j) = 0, no processing is done at position (i, j);

[0049] Otherwise, first determine whether the secret information m(1, k) and m(2, k) are equal. If they are equal, embed the data "1" in the same bit of the third bit plane, otherwise embed the data "0". Among them, m(1, k) and m(2, k) represent the two data in the kth column of the secret information, because the secret information is arranged in 2 rows and n columns, and m(1, k) and m(2, k) represent the two data in the kth column;

[0050] After the third bit plane is embedded with "1", the first bit of secret data is detected. When m(1, i) = 1, the corresponding bit of the first bit plane is embedded with "1". Otherwise, if m(1, i) = 0, the corresponding bit of the first bit plane is embedded with "0".

[0051] Similarly, when the third bit plane is "0", it can be embedded into the first bit plane in the same way;

[0052] Finally, the data embedding of the first and third planes is completed to obtain the stego-image m1, and the embedding capacity N is recorded at the same time.

[0053] Furthermore, in order to reduce the redundant space occupied by the marker bit and to better increase the embedding capacity, the second bit plane is compressed using the BBE compression algorithm, and the excess redundant space can continue to embed secret information;

[0054] Specifically, the mask image I1 is divided into s*s non-overlapping blocks, and the block classification flag type and the embeddable capacity CB are obtained according to the BBE algorithm;

[0055] The redundant part after compression is recorded, which is the space that can be embedded again, and the position that can be embedded again in the block is represented by "1", that is, F(i,j)=1, otherwise, the position that cannot be embedded again is F(i,j)=0. Among them, F represents the bit plane that can be embedded again, and F(i,j)=1 represents the data in the i-th row and j-th column of the bit plane that can be embedded again;

[0056] In order to achieve accurate data extraction, it is necessary to record the data E at the block classification flag in the original block, so it is superimposed on the message as the secret information embedded in the second and fourth planes, and the size length of E is calculated to be k, which can be expressed as message1 = [E, message (1:2, N + 1: end)].

[0057] Repeat the above process until all blocks are compressed, and obtain the data mark bit plane F that can be embedded again and the embedding capacity C that can be increased b .

[0058] Embedding the second and fourth bit planes: Get the position that can be embedded again in the compressed mark bit plane, and embed the data in the same way Figure 2 Data embedding is completed in this way.

[0059] Specifically, the steganographic image G1 embedded in one or three planes is divided into non-overlapping blocks of the same size s*s, and the same method is used. Figure 2 The embedding process embeds into the second and fourth bit planes, records the embedding capacity N1, and changes the marking bit plane from B to F;

[0060] When all blocks are completely embedded with secret information, the secret image G is obtained, and the embedding capacity is recorded as Num=N+N1-k.

[0061] S3: Divide the secret image G into s*s non-overlapping blocks, decrypt the data and recover the steganographic data;

[0062] It should be noted that the secret image G is divided into s*s non-overlapping blocks according to Figure 3 Data decryption and Figure 4 The extraction process recovers the stego data, specifically:

[0063] Determine whether the first bit of data, i.e. m(1, k), is 1;

[0064] Then determine whether the second bit of data m(2, k) is 1. If it is 1, it means the embedded secret data is "11", otherwise it is "00";

[0065] When m(1, k) = 0, if m(2, k) = 1, it means the embedded data is "10", otherwise it is "01";

[0066] By analogy, each column of the data M is restored in a repeated cycle and reorganized to obtain the corresponding original data MB. It can be seen that MB is the combination of the data E and message 1 mentioned in the embedding step above.

[0067] Then, according to the BBE algorithm, the block identification bit data of the second bit plane is used to restore the second bit plane data in the manner of Table 1;

[0068] Table 1: Block type and block flag bit table.

[0069]

[0070]

[0071] Find the corresponding bitmap data from the extracted data MB through the index of the non-embeddable block, and completely extract the bitmap data B embedded in the second bit plane. b ;

[0072] According to the extracted bit mark bitmap B b And the bit mark map B for determining the embedded position f , complete the extraction of data mess1 of the second and fourth planes of the secret image G.

[0073] Bit Marking Figure B b After the complete extraction, according to the embedding method of the first and third planes, the first and third plane information data mess2 are extracted according to the extraction method in the extraction process; the judgment bit mark diagram B b Whether (i,j) is 1. If it is 1, it means that there is data embedded in this position. Otherwise, there is no data embedded.

[0074] When B b When (i, j) = 1, the corresponding bit data m1(i) of the third bit plane of the secret image G is extracted. At this time, the subsequent extraction process is the same as Figure 3 similar;

[0075] By analogy, the first and third plane data are repeatedly extracted cyclically to obtain data mess2, and the final extracted information is message = [mess1, mess2].

[0076] Example 2

[0077] Reference Figures 5 to 9 This is another embodiment of the present invention. Different from the first embodiment, this embodiment provides a verification test of an adaptive high-capacity steganography method for noisy images. In order to verify and illustrate the technical effects used in this method, this embodiment uses traditional technical solutions to conduct comparative tests and example tests with the method of the present invention, and compares the test results by means of scientific argumentation to verify the real effect of this method.

[0078] In the implementation scheme, the parameters are set as follows: the test image size is 512*512, and the pixel value range is [0,255]. P=8 is selected when the test image is decomposed into bit planes. The secret information secret is a randomly generated random sequence, and the block size s=4. In addition, this paper is mainly aimed at remote sensing and noise images in radar, so 6 marine ship detection images are randomly selected from the Sentinel-1 satellite reconnaissance data set SSDD (https: / / zhuanlan.zhihu.com / p / 143794468) as test images. The proposed algorithm is simulated using the software MATLAB 2019a.

[0079] Invisibility Analysis:

[0080] (1) Performance evaluation of the algorithm on images:

[0081] The most basic problem of image steganography is to ensure that the human perception system cannot detect the steganography. For digital images, a series of indicators such as mean squared error (MSE), PSNR and structural similarity index (SSIM) are generally selected to evaluate the image distortion rate before and after steganography. Therefore, the above indicators and payload are selected here to evaluate the impact of the algorithm on the quality of each test image as shown in Table 2.

[0082] Table 2: Performance evaluation table of the algorithm of the present invention on each test image.

[0083] image MSE / % PSNR / % SSIM Payload (a) 14.13 36.63 0.9750 480670 (b) 25.48 34.07 0.9422 733004 (c) 16.68 35.91 0.9779 549728 (d) 21.19 34.87 0.9885 649176 (e) 18.15 35.54 0.9830 582088 (f) 14.95 36.37 0.9934 508438

[0084] From Table 2, we can see that the mean square error MSE is almost close to 0, indicating that the original image and the steganographic image fit well and the data embedding is relatively stable; from the peak signal-to-noise ratio PSNR and the payload value, it can be confirmed that the present invention can meet the high-capacity embedding while ensuring good image quality; and the structural similarity index SSIM value is almost close to 1, which means that the algorithm in this paper has little effect on the image quality.

[0085] (2) Image visual invisibility analysis

[0086] The present invention compares the performance of 6 test images. The effective load when fully embedded is shown in Table 3. The experimental results are shown in Table 3. Figure 5 shown.

[0087] Table 3: Statistics of effective loads of each algorithm when fully embedded.

[0088] picture (a) (b) (c) (d) (e) (f) proposed 580670 733004 549728 649176 582088 508438

[0089] According to the experimental results Figure 5 It can be seen that the effective load number of the method of the present invention is higher when it is fully embedded, and the results show that the performance of this algorithm is better, that is, when it is fully embedded, the proposed algorithm has the best visual effect. Figure 5 It can be seen from the curve that the visual quality values ​​of the six test images do not change much, indicating that the embedding effect of the present invention on different test images is relatively stable.

[0090] (3) Image statistical invisibility analysis:

[0091] At present, the existing algorithms for steganalysis of spatial domain image steganography technology generally include RS (Regular Singular) algorithm, Chi-square steganalysis algorithm, information estimation method and GPC (Gray-Level Plane Crossing) algorithm, etc. Among them, RS algorithm is most accurate when the embedding rate is 0% and 100%, and there is a 10% error in others. Chi-square steganalysis algorithm generally detects embedding when the embedding rate is about 70%, and the result is 0 when it is lower than this number; the information estimation method is only accurate at low embedding rates, and the errors in others are large; the GPC analysis method is more accurate when the steganography rate is large, and the curve is also stable; therefore, the GPC analysis method is used here to detect each test image, and the results are as follows Figure 6 As shown, Figure 6 It can be seen that in addition to achieving a high embedding rate, the algorithm also has a low detection result using the GPC steganalysis method, indicating that the present invention has good statistical invisibility while achieving a high embedding rate.

[0092] Embedding capacity analysis: In order to more intuitively show the changes in the embedding capacity and PSNR value of the test image, different amounts of information are embedded in the test image, and the PSNR values ​​when different amounts of information are embedded are calculated. The results are as follows: Figure 7 As shown, from Figure 7 The results show that the PSNR value of the present invention decreases as the embedding capacity increases, but the PSNR value is higher when the embedding rate is small and when the embedding rate is full. This is because in addition to using a genetic algorithm to adaptively select embedding points for noisy images to embed the encryption algorithm of the cross-bit plane, the present invention also uses the BBE algorithm to compress the mark bit to increase the embedding capacity, so that the image can meet the high-capacity embedding while ensuring a higher visual effect.

[0093] Robustness analysis:

[0094] (1) Data anti-interference detection

[0095] In practical applications, some data loss or other data attacks are inevitable. To test the robustness of the present invention, a grayscale image of size 128*128 and the first test image are used as the secret image and the carrier image for robustness experiments. Figure 8 The following are attacks on the secret image, such as MSB loss, LSB loss, random 50*50 data shearing, and median filtering. Fig. 9 represents the secret image extracted after the attack. Fig. 9 It can be seen that the present invention can still completely extract the hidden information after the above attacks. Although there are differences in data extraction under certain attacks, the secret and important information is still clearly visible, which proves that the algorithm in this paper has strong anti-interference ability and good robustness.

[0096] (2) Image anti-interference detection

[0097] The secret image was sharpened and rotated 5 degrees at the same embedding rate. The results are shown in Table 4. The evaluation standard NCC is an algorithm based on statistics to calculate the correlation between two sets of sample data. Its value range is [-1,1]. For images, if the two images match each other, the NCC value is 1, if they are completely different, it is -1. The calculation formula is as follows:

[0098]

[0099] Where f is the gray value, μ is the average value of all pixels, and σ is the standard deviation. Assuming t represents the pixel value of the contrast image, its calculation formula is as follows:

[0100]

[0101] Where n is the total number of pixels in the contrast image and n-1 is the degree of freedom.

[0102] Table 4: Robustness test table of the test image against different attacks.

[0103]

[0104] Table 4 shows the NCC values ​​of the test images (a)-(f) after the above attacks. It can be seen that the data correlation NCC values ​​of the original image and the stego image are close to 1, which shows that the present invention has good anti-interference ability.

[0105] The specific results of simulation experiments and performance comparison experiments show that the present invention has good effects in terms of image embedding rate and PSNR results; experimental analysis can also confirm that the present invention has good resistance to existing good steganalysis algorithms, and has good robustness to different attacks, so it can be used in radar, remote sensing image data transmission, information steganography and other fields.

[0106] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. An adaptive high-capacity steganographic method for noisy images, characterized in that: include: Input an image and preprocess the image; Data embedding is performed on the preprocessed image to obtain the secret image G; Divide the secret image G into s*s non-overlapping blocks, and perform data decryption and restore the steganographic data; The pre-processing step comprises: The BM3D algorithm is used to filter and denoise the areas with high noise corresponding to the image frequency; Compare the data obtained after filtering by the BM3D algorithm with the optimal embedding threshold T obtained by the genetic algorithm, find out the points where the noise value is higher than the preset value and generate the global candidate steganographic position, and obtain the marked bit plane B; The marking bit plane B replaces the second bit plane of the image pixel in LSB mode to obtain a masked image I1; The input image I is decomposed into bit planes to obtain P binary images, where the calculation formula for the corresponding bit of the pixel in the kth bit plane is: Among them, i, j represent the corresponding position of the pixel, i-th row, j-th column, mod2 represents the remainder when divided by 2, I (k) (i, j) represents the pixel value of the i-th row and j-th column in the k-th bit plane of plane I; The preprocessed image is embedded with data to obtain the secret image G, which includes: Performing bit plane decomposition on the mask image I1, reorganizing the secret information into a matrix sequence message, and extracting the second bit plane of the mask image I1, namely, the marking bit plane B; If B(i, j) = 0, no processing is done at position (i, j); Otherwise, first determine whether the secret information m(1, k) and m(2, k) are equal. If they are equal, embed the data "1" in the same bit of the third bit plane, otherwise embed the data "0"; After the third bit plane is embedded with "1", the first bit of secret data is detected. When m(1, i) = 1, the corresponding bit of the first bit plane is embedded with "1". Otherwise, if m(1, i) = 0, the corresponding bit of the first bit plane is embedded with "0". Similarly, when the third bit plane is "0", it can be embedded into the first bit plane in the same way; Finally, the data embedding of the first and third planes is completed to obtain the steganographic image G1, and the embedding capacity N is recorded at the same time.

2. The adaptive high-capacity steganographic method for noisy images as claimed in claim 1, characterized in that: Also includes, Divide the mask image I1 into s*s non-overlapping blocks, and obtain the block classification mark type and the embeddable capacity CB according to the BBE algorithm; The redundant part after compression is recorded, which is the space that can be embedded again, and the position that can be embedded again in the block is represented by "1", that is, F(i, j) = 1, otherwise, the position that cannot be embedded again is F(i, j) = 0; Record the data E at the block classification flag in the original block, add it to the message as the secret information embedded in the second and fourth bit planes, and calculate the size of E to be k, expressed as message1 = [E, message (1: 2, N + 1: end)]; Repeat the above process until all blocks are compressed, and obtain the data mark bit plane F that can be embedded again and the embedding capacity C that can be increased b .

3. The adaptive high-capacity steganographic method for noisy images as claimed in claim 2, characterized in that: Also includes, Second and fourth plane embedding: Divide the stego-image G1 after embedding in the first and third bit planes into non-overlapping blocks of the same size s*s, and embed them into the second and fourth bit planes according to the embedding process, record the embedding capacity N1, and mark the bit plane from B to F; When all blocks are completely embedded with secret information, the secret image G is obtained, and the embedding capacity is recorded as Num=N+N1-k.

4. The adaptive high-capacity steganographic method for noisy images as claimed in claim 3, characterized in that: Dividing the secret image G into s*s non-overlapping blocks, and performing data decryption and recovering the steganographic data includes: Determine whether the first bit of data, i.e. m(1, k), is 1; Then determine whether the second bit of data m(2, k) is 1. If it is 1, it means the embedded secret data is "11", otherwise it is "00"; When m(1, k) = 0, if m(2, k) = 1, it means the embedded data is "10", otherwise it is "01"; By analogy, each column of the data M is restored in a repeated cycle and reorganized to obtain the corresponding original data MB. It can be seen that MB is the combination of the data E and message 1 mentioned in the embedding step above.

5. The adaptive high-capacity steganographic method for noisy images as claimed in claim 4, characterized in that: Also includes, According to the BBE algorithm, the block identification bit data of the second bit plane is used to restore the second bit plane data; Find the corresponding bitmap data from the extracted data MB through the index of the non-embeddable block, and completely extract the bitmap data B embedded in the second bit plane. b ; According to the extracted bit mark bitmap B b And the bit mark map B for determining the embedded position f , complete the extraction of data mess1 of the second and fourth planes of the secret image G.

6. The adaptive high-capacity steganographic method for noisy images as claimed in claim 5, characterized in that: Also includes, The bitmap B b After the complete extraction, according to the embedding method of the first and third planes, the first and third plane information data mess2 are extracted according to the extraction method in the extraction process; the bit mark map B is judged b Whether (i, j) is 1. If it is 1, it means that there is data embedded in this position, otherwise, there is no data embedded; When B b When (i, j) = 1, extract the corresponding bit data m1(i) of the third bit plane of the secret image G; By analogy, the first and third plane data are repeatedly extracted cyclically to obtain data mess2, and the final extracted information is message = [mess1, mess2].

Images