CN108810555B - Thumbnail image privacy protection method based on image compression and elastic resolution - Google Patents

Abstract

A thumbnail image privacy protection method based on image compression and elastic resolution; by dividing the original image into blocks, and subdividing each block into sub-blocks, according to the requirements of thumbnail resolution, each block is divided into sub-blocks. The sub-blocks are randomly scrambled to a corresponding degree; combined with JPEG compression and cryptography, the DC and AC coefficients in the 8 × 8 blocks subjected to DCT transform are encrypted to different degrees, and the thumbnails of the public part and the private parts are obtained. The data is completely encrypted; the thumbnail image of the public part in the present invention can be used for the picture preview of the network application, and at the same time, a certain degree of privacy protection is provided. Using image compression technology to reduce the storage overhead of the encrypted image as a whole, especially the public part, can greatly save the bandwidth. The invention also provides thumbnail privacy protection of elastic resolution, the security of the image is affected by the resolution, and the lower the resolution is controlled, the higher the security of the image.

Description

Thumbnail image privacy protection method based on image compression and elastic resolution

【Technical field】

The invention belongs to the technical fields of digital image encryption, digital image privacy protection and information security, and in particular relates to a thumbnail image privacy protection method based on image compression and elastic resolution.

【Background technique】

Compared with sound and text information, digital image overcomes the difficulty of large storage space and can convey more information because of its vivid content and image, thus becoming one of the most important ways to spread information [1]. With the rapid development of Internet applications, various cloud storage devices have begun to be used as management tools for digital images. It can easily share large files asynchronously without point-to-point transfer. Even if the mobile device is lost or damaged, it can be downloaded from the cloud. The low cost of accessing and restoring data on the cloud saves users a lot of money for transferring and managing data. As a result, almost one billion new photos are uploaded to the cloud application every week.

However, the convenience brought by cloud storage also poses a threat to the privacy of users, because uploading any meaningful image to a cloud application is equivalent to making a backup on an untrusted third-party server, so the privacy of the user is very important. may be leaked. The reasons for this hidden danger are: first, the service provider will leak the user's photo to the partner for some commercial interests; second, the user's account may be attacked by attackers, which will make the user's photo The photos fell into the hands of the attackers, resulting in the disclosure of the user's face information, life information, etc.

The solution to this problem is to implement "end-to-end" encryption, that is, the sender needs to encrypt the image before uploading it to the server, so that the image obtained by the server is the encrypted image, and there is no authorized third party. Without the key, the ciphertext image cannot be restored to the original image, which greatly improves the security of the image content from being stolen, and plays a good role in protecting the digital image involving personal privacy. Technologies in this area include image encryption technology based on matrix transformation pixel replacement, image encryption technology based on secret segmentation and sharing, and image encryption technology based on modern cryptosystems [2].

However, although the traditional classic image encryption algorithm has solved the problem of image privacy protection well, it is very difficult to retrieve images due to incompatible formats or extremely confusing encrypted images, and users cannot preview images online. References [3], [4] solved the problem of format incompatibility, but still did not solve the problem of image preview. In [5], Wright et al proposed a thumbnail image privacy protection algorithm at the Information Hiding and Multimedia Security Conference (TPE), choose to provide a low-resolution JPEG thumbnail for each image, which provides a compatible image format, hides the details of the image, protects the user's privacy to a certain extent, and can preview online .

However, whether it is the traditional image encryption algorithm or the latest thumbnail image privacy protection algorithm, the spatial correlation of the image is changed, which greatly increases the storage cost of the encrypted image file, and there is obvious block effect after the image reconstruction of the TPE algorithm. , the visual quality cannot reach the effect accepted by the human eye. The thumbnail privacy protection algorithm proposed in Reference [6] has improved image security, file size and image reconstruction quality, but the overall effect is still unsatisfactory.

references

[1] Gonzalez, Rafael C., and Richard E. Woods. "Digital image processing." (2012).

[2] Li Changgang, Han Zhengzhi, and Zhang Haoran. "Overview of Image Encryption Technology." Computer Research and Development 39.10 (2002): 1317-1324.

[3] Tierney, Matt, et al. "Cryptagram: Photo privacy for online socialmedia." Proceedings of the first ACM conference on Online social networks. ACM, 2013.

[4] Ra, Moo-Ryong, Ramesh Govindan, and Antonio Ortega. "P3: TowardPrivacy-Preserving Photo Sharing."Nsdi.2013.

[5]Wright,Charles V.,Wu-chi Feng,and Feng Liu."Thumbnail-preserving encryption for JPEG."Proceedings of the 3rd ACM Workshop on InformationHiding and Multimedia Security.ACM,2015.

[6] Marohn, Byron, et al. "Approximate Thumbnail Preserving Encryption." Proceedings of the 2017on MultimediaPrivacy and Security.ACM,2017.

[Content of the invention]

The purpose of the invention is to solve the following key technical problems:

First, an image encryption method for saving the original image format is proposed to solve the problem of incompatible encrypted image formats.

Second, a method for privacy protection of thumbnail images is proposed. Encrypted images can not only protect the privacy of users, but also can be previewed online through thumbnail images.

Thirdly, an image encryption method based on image compression is proposed, so that the encrypted image can save the file storage space and transmission bandwidth as much as possible.

Fourth, an elastic resolution thumbnail image encryption method is proposed, and the thumbnail resolution of the image can be controlled by the user.

In order to solve the above problem, the present invention reads the format of the original image, and still stores the format of the original image after performing the image encryption process, so that the usability of the encrypted image is stronger.

In order to solve the above problem, the present invention proposes to create a low-resolution thumbnail of the original image. The details of the image are discarded in the thumbnail, and only the rough outline of the image is preserved, and it is difficult to restore the approximate original image from the thumbnail. This method not only protects the private information in the image, but also enables online preview by encrypting the image.

In order to solve the above problem, the present invention uses image compression technology to control the extra storage overhead brought by image encryption, and removes image redundancy through reasonable image quantization to reduce the size of the original file as much as possible. And combined with the characteristics of image compression technology, the encrypted image is divided into public part and private part. The public part is a small thumbnail, and only the public part is loaded when previewing online through the thumbnail, which can greatly save bandwidth.

In order to solve the above problem, the present invention provides a thumbnail image encryption technology with elastic resolution, and the detail exposure degree of the image is controlled by the user. At the same time, the security of images depends not only on the security of the algorithm, but also on the size of the thumbnail resolution. The higher the resolution of the image, the less detail is exposed in the image, and the higher the security of the image.

In order to solve the above problems, the present invention also provides a prototype system for image storage with thumbnail encryption. Supports flexible selection of image encryption levels and corresponding resolution levels for uploading images, which will be decrypted into high-definition images on the terminal during download.

Technical scheme of the present invention:

A thumbnail image privacy protection method based on image compression and elastic resolution, the method includes a thumbnail encryption method and a thumbnail decryption method;

1. Thumbnail encryption method

1. Divide the image into blocks in the spatial domain: first read the image information in the bitmap image file, assuming that the image size of the image is N×N, divide the original image according to n×n, n<N, The same below;

1.1. Divide each n×n block twice, the size of each sub-block is b×b, b<n, the same below;

1.2. Randomly scramble several sub-blocks in each n×n block. Considering the nature of JPEG compression, in order to better obtain image restoration quality, n mod8≡0, b mod8≡0, and b involved in scrambling are adopted. The number of ×b sub-blocks is determined by the sub-key k ₀ , so as to flexibly control the resolution of the thumbnail image, and the random positions of the sub-blocks involved in scrambling are generated by a pseudo-random number generator;

Assuming that each N×M block contains m sub-blocks of b×b, the scrambling degree of the image is divided into 5 levels: when level=0, the image is neither scrambled nor compressed, and the original image is completely preserved; level= 1, m/4 sub-blocks participate in scrambling; level=2, m/2 sub-blocks participate in scrambling; level=3, 3m/4 sub-blocks participate in scrambling; level=4, m sub-blocks participate in scrambling ; level will be used as the "seed" of the pseudo-random number generator to determine the random scrambling position of the sub-block;

2. Using the JPEG compression method to transform the image into the frequency domain through the forward discrete cosine transform (DCT), the following rules should be followed for the DCT transformation of the image here:

2.1. Subtract 128 from each pixel data of the image, so that the range of pixel values falls within the range of -128 to 127, which can greatly reduce the probability that the decimal number after DCT transformation is 3 bits, which is convenient when using VIL encoding, reduce the number of bits encoded;

2.2. Combined with the nature of JPEG compression, apply DCT transform to each 8×8 image block to obtain 1 DC coefficient and 63 AC coefficients, and at the same time, the DC coefficients are called low-frequency coefficients, and the AC coefficients are called high-frequency coefficients;

`3. Quantize the DCT transform result in step 2, that is, divide every 8×8 block of the image by the default quantization table in JPEG compression, and the result is rounded to the nearest integer; in this way, most of the high-frequency coefficients in the DCT coefficients change. If it is "0", some coefficients that are not set to "0" will also lose some accuracy; the purpose of quantization is to filter the high-frequency coefficients in the image, and the main information of the image is concentrated in the low-frequency part of the image, which can play a role in The effect of compression will not have a large impact on the main information of the image.

4. Perform zigzag scanning on every 8×8 DCT coefficients quantized in step 3, so that a 1×64 one-dimensional sequence will be obtained for the 8×8 blocks of DCT coefficients;

4.1. Separate the high-frequency coefficient AC from the low-frequency coefficient DC, store the high-frequency coefficient as a private part, and replace the position of the high-frequency coefficient with "0";

4.2. Perform zero-run coding and Huffman coding on the separated high-frequency coefficients, further compress the sequence coding length of the data, and obtain binary sequences w ₀ , w ₁ , . . . , _wi ;

4.3. Perform stream cipher encryption on the high-frequency coefficients after step 4.2;

The specific method of stream cipher encryption is: use the subkey k ₁ to generate a binary key stream r ₀ , r ₁ ,..., _ri through the stream cipher generator, and press the binary _wi sequence in step 4.2 according to The encrypted w _i ' sequence is obtained by the following formula:

w _i ' will be saved as a private part;

Fifth, the coefficients after the high-frequency coefficients are set to "0" are subjected to izigzag scanning and inverse quantization. The result of inverse quantization discards the details of the image and only retains the rough features of the image;

6. Inverse discrete cosine transform (IDCT) is performed on the DCT coefficients processed in step 5 to the spatial domain;

7. In step 2.1, subtract 128 from all pixel values, and add 128 here to restore the pixel values to the range of 0-255, so that the corresponding scrambling level thumbnails are obtained;

8. Encrypt the thumbnail image in step 7 again in blocks, and the size of each block is 8×8;

8.1. At this time, the pixel value in each block of the thumbnail is the average value of the block, so it is necessary to randomly scramble the lower six bits of the pixel in each block, so that the pixel value of the block is not an exact average value, but close to the average;

8.2. In order to ensure the compression effect of the image, the same sub-key k ₂ is used for the 64 pixels in the same block to scrambling the pixel values by six positions lower;

Although the sub-blocks in each block are scrambled to a certain extent in step 1, it brings a certain degree of difficulty for the attacker to brute force the image. However, in steps 3-4.1, DCT transform is performed on each 8×8 block, and all ACs are set to “0”. Due to the characteristics of DCT transform, after the image undergoes IDCT transform, each 8×8 block is The value of the middle pixel is the average value of the block, which accurately exposes the average value of each 8x8 block after inter-block scrambling. Therefore, in step 8.2, the subkey k ₂ is taken as the "seed" of the random number generator, and the lower six bits of each pixel value in each 8×8 block are randomly scrambled; Correlation, easy to compress, the subkey in each block is the same, which not only hides the average value of the block to a certain extent, but also makes each block have a certain degree of similarity with the original image.

2. Thumbnail decryption method

9. Divide the encrypted thumbnail image into 8×8 blocks, and use the sub-key k ₂ to restore and decrypt the lower six bits of the pixel value in each block;

10. Decrypt the high frequency coefficient AC separated in step 4 according to the following formula according to the subkey k ₁ ;

11. Perform DCT transformation and quantization on the thumbnail as in steps 2-3;

11.1. The high-frequency coefficients in the DCT coefficients are replaced by the high-frequency coefficients decrypted in step 10;

11.2. Perform inverse quantization and IDCT transformation on the thumbnails to obtain slightly higher resolution thumbnails;

12. Use the subkey k ₀ to restore the scrambled block to the position of the original image, and restore it to the original high-definition image.

It can be found that the finally restored image is not the original image, because the lossy mode in the compression algorithm in the encryption algorithm of the present invention performs a quantization operation on the transformed coefficients, in other words, the algorithm of the present invention is lossy, But the final image restoration quality is acceptable.

Advantages and beneficial effects of the present invention:

1. The present invention provides an elastic resolution privacy protection scheme, and in practical applications, the user can control the resolution of the image by himself. At the same time, the flexible resolution also flexibly controls the security of thumbnails. For images that do not require very high security, a higher resolution can be used to facilitate online preview; images with strong security requirements use a lower resolution. Facilitate image privacy protection;

2. In addition, the public part and the private part of the image are stored separately, and neither party who saves the two parts can decrypt the image alone, which makes it more difficult for the attacker to crack the encrypted image;

3. Compared with the previous thumbnail image privacy protection scheme, the present invention combines the JPEG image compression algorithm, which not only removes the detailed information part in the image, but also reduces the storage overhead after image encryption more efficiently. The resolution is flexible, so when the selected resolution is higher, the advantage of saving the storage overhead of the encrypted image will be more obvious, and the size of the encrypted image in both high resolution and low resolution is significantly smaller than the original image. , which was not achieved by previous schemes;

4. In the algorithm of the present invention, in order to reduce the storage overhead, DCT transform is used, and the coefficients are quantized, which makes the original image irretrievably lose some information, in other words, the solution of the present invention is lossy, However, on the premise of ensuring image security and compression efficiency, it provides acceptable image restoration quality.

The preferred embodiments of the present invention will be described in detail below in conjunction with the accompanying drawings and specific embodiments, and these and other advantages of the present invention will be more apparent.

【Description of drawings】

Fig. 1 provides the example diagram for the image block of N*N size in the present invention;

2 is a flow chart of M200 for performing elastic resolution thumbnail encryption on an image according to an embodiment of the present invention;

FIG. 3 shows the M300 flow chart for dividing and scrambling sub-blocks in the present invention.

4 is an example of random scrambling between sub-blocks when the scrambling level is 1-4 in 32×32 blocks according to an embodiment of the present invention;

Figure 5 shows the effect of scrambling between image sub-blocks under different scrambling levels;

Figure 6 shows a schematic diagram of setting the AC coefficients to 0 after DCT transformation, where the minimum unit block represents one pixel.

Fig. 7 is a specific example in which AC coefficients are set to 0 after DCT transformation after inter-block scrambling;

Fig. 8 provides the schematic diagram of the random scrambling of pixel values in the present invention by the lower six bits;

Figure 9 shows the comparison between the public part of the encrypted image and the size of the original image file, and the data unit is KB;

Figure 10 shows the comparison of the size of the encrypted image as a whole and the original image file, the data unit is KB;

Figure 11 lists the results of edge detection on thumbnails at different scrambling levels;

Figure 12 is a flowchart of the thumbnail image decryption algorithm M1200;

Figure 13 uses the PSNR value to measure the reconstruction quality of the decrypted image, and the data unit is db;

Figure 14 shows the image reconstruction effects of three different resolution images.

【Detailed ways】

Specific embodiments of the present invention will be described below with reference to the accompanying drawings.

For the sake of convenience, FIG. 1 shows an example diagram of the block in the present invention. For an N×N image, it is divided into several n×n blocks, and each n×n block is divided twice, and divided into b Example diagram of ×b sub-blocks.

Example 1:

Figure 2 shows a flow chart of an M200 embodiment of the algorithm encryption process of the present invention.

As shown in FIG. 2, the thumbnail encryption process starts at step S210.

S220 divides the image into 64 blocks of 32×32 for an image with an image size of 256×256, and divides each 32×32 block into 8×8 sub-blocks, and S230 divides the sub-blocks in each block Random scrambling is performed. In this process, the degree of scrambling between sub-blocks is determined by the level of scrambling. Different levels have different scrambling degrees between sub-blocks, which directly determines the resolution of the final image. The randomness of the sub-blocks The scrambling position is determined by the subkey k ₀ (for the detailed inter-subblock scrambling process, see M300).

Figure 4 shows a specific example of random scrambling between blocks. For a 32×32 image block, it is divided twice, and the size of each sub-block is 8×8, so the 32×32 block contains 16 sub-blocks. In each unit in 4, the left picture is the original block, the right picture is the result of random scrambling of the left picture sub-block, and the shaded block is the randomly selected block that needs to be scrambled. According to the definition in step 1.2 of the thumbnail encryption algorithm, when the scrambling level is 0, no scrambling is performed, and when the scrambling level is 1, 4 blocks that need to be scrambled are randomly selected, ..., and so on, the scrambling level is 4, all 16 sub-blocks are randomly scrambled, as shown in Figure 4.

Figure 5 shows the scrambling effect diagrams between image sub-blocks under different scrambling levels in turn. It can be seen that the effect of this local block scrambling still shows the general outline of the image. When level=0, the image remains the original image, and as the scrambling level increases, the image becomes more and more blurred, which lays the foundation for the elastic resolution thumbnail in the algorithm of the present invention.

S240 in FIG. 2 is to perform DCT transformation on every 8×8 block of the image according to the method in JPEG compression, and quantize it to obtain high-frequency coefficients DC and low-frequency coefficients AC, and keep DC and separate AC.

S250 in FIG. 2 sets the part of the original AC coefficient in S240 to 0, and performs inverse quantization and IDCT transformation to obtain a thumbnail image with a lower resolution.

Figure 6 shows a schematic diagram of setting the AC coefficients to 0 after DCT transformation for each 8×8 block. The upper left corner of the left picture is the DC coefficient, and the rest are AC coefficients. The right picture shows the original AC coefficient after the AC is separated. position is set to 0.

Figure 7 is a schematic diagram of an image after DCT transformation is performed on the scrambled image and AC is set to 0 according to the scrambling level of 3. DC concentrates the most important energy in the image, so when AC is separated, most of the information in the image is saved with less detail, so the resolution of the image will be lower.

S260 in Figure 2 is to use zigzag to scan the DCT coefficients of S240 to obtain a one-dimensional block sequence, separate the AC coefficients, and use the subkey k ₁ to perform stream cipher encryption and save it as a private part.

S270 in Fig. 2 re-divides the thumbnail image generated by S250 into blocks, the size of each block is 8×8, and the lower six bits of the pixel value in each block are scrambled using the same seed key, and this part is used as the public key. Partially saved.

Figure 8 shows a schematic diagram of scrambling the pixel values of 256 gray levels in the lower six positions. After scrambling, the pixel value is no longer the exact average value of the original image, but because the upper 2 bits have not changed, the pixel value is still close to on average.

The thumbnail encryption process ends at S280;

Thus, the encrypted image of the elastic resolution thumbnail of the public part and the encrypted sequence of the private part are obtained. In order to measure the cost of consuming storage space after image encryption, the invention compares the storage space consumed by ciphertext and plaintext.

Figure 9 compares the file size (unit: KB) between the public part of the encrypted image and the plaintext under different scrambling levels. When level=0, it means the original image, and its file size is 39.7KB. When the scrambling level is above 1, the public part is about 0.5KB. Compared with the original image, the thumbnail part is greatly reduced, which will greatly save the storage overhead of the public part of the server, and at the same time, it can save the bandwidth to a greater extent and improve the performance of the server.

Figure 10 shows the comparison of the file size of the encrypted file (public part + private part) and the original image. It can be seen from the figure that the solution of the present invention not only has no additional overhead, but also reduces the storage space of the original image. A lot, which is a big improvement over the previous thumbnail image privacy protection algorithm.

Edge detection is a common method for attacking privacy nowadays, because the most important information in the image is contained in the boundary part of the image. The invention uses the most commonly used Canny operator edge detection method in the MATLAB toolbox to evaluate the security of thumbnail images.

Figure 11 shows the original image and different levels of level0-level4, some feature points can be detected using edge detection, where level0 and level1-level4 represent the results of edge detection on the original image and thumbnails of different resolutions, respectively. It can be seen that with the increase of the scrambling level, the effect of image edge detection is getting worse and worse. For experts in the field of computer vision or signal processing, because they also collect various data sets, they may be able to learn from these exposures. The feature points analyze some information, but if the picture is not in any dataset, then it is impossible to discern what is in the image, or for a layman, no information can be obtained from the results in the picture.

FIG. 12 is a flowchart of a method M1200 for decrypting a thumbnail image. _The method starts from S1210, and in step S1220, the lower six bits of all pixel values are restored through the subkey k2. S1230 will use the subkey k ₁ to decrypt the private part data to obtain the decrypted AC sequence.

In step S1240, DCT transform and quantization are applied to the thumbnails according to JPEG compression. In step S1250, the AC coefficients decrypted in S1230 are used to replace the AC coefficients obtained in step S1240, and inverse quantization and IDCT transformation are performed to obtain the restored image.

S1260 performs an inverse scrambling operation on the image according to the subkey k ₀ to obtain a restored high-definition original image, and the thumbnail image decryption algorithm ends at S1270.

It should be noted that the final restored image is not the original image, because the lossy mode in JPEG compression is used in the encryption algorithm, and the coefficients after DCT transformation are quantized, which loses part of the information of the original image, in other words In other words, the algorithm of the present invention is lossy, but the final image restoration quality is acceptable.

Figure 13 lists the restoration quality data of the image of the present invention. Studies have shown that when the PSNR value of the reconstructed image and the original image is above 28db, the image quality acceptable to the human eye is achieved.

It can be seen from Figure 13 that the restoration quality of the image is within the acceptable range for human eyes, and the reconstruction quality of the image hardly changes due to the change of the block size and scrambling level, which is due to the loss of the image It is only reflected in the process of quantization after DCT transformation. Although the present invention divides different blocks, the sub-blocks in each block are multiples of 8×8, and the pixel position in each sub-block does not change, so the image Neither the size of the block nor the degree of scrambling between sub-blocks is a factor in image quality.

FIG. 14 is a schematic diagram of the image after image reconstruction when the scrambling level level=4 for images of different resolutions. The image resolutions of the first to third columns are 256×256, 512×512 and 1024×1024, respectively. It is not difficult to find that when the resolution of the image is higher, no trace of image loss can be found from the perspective of the human eye. Therefore, the reconstruction quality effect of the algorithm of the present invention is very considerable.

Claims (1)

1. A thumbnail image privacy protection method based on image compression and elastic resolution comprises a thumbnail encryption method and a thumbnail decryption method;

first part, thumbnail encryption method

In the thumbnail image encryption process, the image is finally divided into a public part and a private part, the public part is a low-resolution thumbnail, and the private part is a fully encrypted binary sequence, and the specific contents are as follows:

1, partitioning an image in a spatial domain: reading image information in a bitmap image file, and partitioning an original image according to N multiplied by N, wherein N is less than N, and the image size of the image is assumed to be N multiplied by N;

1.1, dividing each block with the size of n multiplied by n for the second time, wherein the size of each sub-block is b multiplied by b, and b is less than n;

1.2, randomly scrambling a plurality of sub-blocks in each n × n block, taking the properties of JPEG compression into consideration, and adopting n mod8 ≡ 0 and b mod8 ≡ 0 in order to better obtain the image recovery quality, wherein the number of the sub-blocks b × b participating in scrambling is determined by a sub-key k₀Deciding to elastically control a resolution of the thumbnail image, and a random position of the sub-block participating in the scrambling is generated by a pseudo random number generator;

assuming that each N × N block contains m b × b sub-blocks, the scrambling level of the image is divided into 5 levels: when the level is 0, the image is not scrambled and is not compressed, and the original image is completely reserved; level is 1, and m/4 subblocks participate in scrambling; 2, m/2 subblocks participate in scrambling; sub-blocks with the level of 3, 3m/4 participate in scrambling; level is 4, and m subblocks participate in scrambling; the level is used as the seed of a pseudo-random number generator to determine the random scrambling position of the subblocks;

2, taking reference to the JPEG compression method, the image is DCT-transformed into the frequency domain through forward discrete cosine transform, wherein the DCT-transformation of the image follows the following rules:

2.1, subtracting 128 from each pixel data of the image to make the range of the pixel value fall into the range between-128 and 127, thus greatly reducing the probability that the decimal number is 3 bits after DCT conversion and being convenient for reducing the coded digit when VIL coding is used;

2.2, by combining the properties of JPEG compression, applying DCT (discrete cosine transformation) to each 8 x 8 image block to obtain 1 DC coefficient and 63 AC coefficients, and simultaneously, calling the DC coefficient as a low-frequency coefficient and the AC coefficient as a high-frequency coefficient;

3, quantizing the DCT transform result in the step 2, namely dividing each 8 multiplied by 8 block of the image by a default quantization table in JPEG compression, and rounding the result to a reserved integer; thus, most of the high frequency coefficients in the DCT coefficients become "0", and some of the coefficients that are not set to "0" will also suffer some loss in accuracy; the quantization aims at filtering high-frequency coefficients in the image, and the main information of the image is concentrated in the low-frequency part of the image, so that the compression effect can be achieved, and the main information of the image cannot be greatly influenced;

4, zigzag scanning is carried out on each 8 × 8 DCT coefficient quantized in the step 3, so that a 1 × 64 one-dimensional sequence is obtained for the 8 × 8 DCT coefficients;

4.1, separating a high-frequency coefficient AC from a low-frequency coefficient DC, storing the high-frequency coefficient as a private part, and replacing the position of the high-frequency coefficient with '0';

4.2, carrying out zero run coding and Huffman coding on the separated high-frequency coefficient, further compressing the sequence coding length of the data to obtain a binary sequence w₀，w₁，...，w_i；

4.3, carrying out stream cipher encryption on the high-frequency coefficient in the step 4.2;

the specific method for encrypting the stream cipher comprises the following steps: using subkey k₁Generating a binary keystream r by a stream cipher generator₀，r₁，...，r_iAnd for the binary value w in the 4.2 step_iThe sequence is encrypted as follows to obtain w_i' sequence:

w_i' will be saved as a private part;

5, carrying out the alizgzag scanning and inverse quantization on the coefficient with the high-frequency coefficient set to be 0, discarding the detailed part of the image as the result of inverse quantization, and only keeping the rough characteristic of the image;

6, carrying out Inverse Discrete Cosine Transform (IDCT) on the DCT coefficient processed in the step 5 to transform the DCT coefficient into a spatial domain;

step 7, subtracting 128 from all pixel values in step 2.1, adding 128, and restoring the pixel values to the range of 0-255, thus obtaining the thumbnail of the corresponding scrambling level;

8, carrying out block encryption on the thumbnail in the step 7 again, wherein the size of each block is 8 multiplied by 8;

8.1, when the pixel value in each block of the thumbnail is the average value of the block, the lower six bits of the pixel in each block need to be randomly scrambled, so that the pixel value of the block is not the exact average value but is close to the average value;

8.2, to ensure the compression effect of the image, the same sub-key k is used for 64 pixels in the same block₂Carrying out pixel value lower six-position scrambling;

although the sub-blocks in each block are scrambled to a certain degree in the step 1, which brings a certain difficulty for an attacker to violently crack the image, in the steps 2 to 4.1, DCT transformation is carried out on each 8 × 8 block, all ACs are set to be 0, and due to the characteristic of DCT transformation, after the IDCT transformation is carried out on the image, the value of each pixel point in each 8 × 8 block is the average value of the block, so that the average value of each 8 × 8 block after the block is scrambled is accurately exposed, and therefore, a sub-key k is adopted in the step 8.2₂In order to keep the correlation of the pixels in the blocks and facilitate compression, the subkeys in each block are the same, thus hiding the average value of the blocks to a certain extent and ensuring that each block has certain similarity with the original image;

second part, abbreviated diagram encryption method

9 th, the encrypted thumbnail image is blocked as 8 × 8 and a subkey k is used₂Carrying out lower six-bit reduction decryption operation on the pixel value in each block;

10 th, according to the subkey k₁Decrypting the high-frequency coefficient AC separated in the step 4 according to the following formula;

11, performing DCT transformation and quantization on the thumbnail as in steps 2-3;

11.1, replacing the high-frequency coefficient in the DCT coefficient by the decrypted high-frequency coefficient in the step 10;

11.2, carrying out inverse quantization and IDCT transformation on the thumbnail to obtain the thumbnail with higher resolution than the thumbnail in the step 8.2;

12 th, use of subkey k₀And restoring the scrambled blocks to the position of the original image, and restoring the original image with high definition.

Images