CN116320201A

CN116320201A - Image two-dimensional reversible information hiding method and system based on double ordering

Info

Publication number: CN116320201A
Application number: CN202310182346.7A
Authority: CN
Inventors: 马宾; 王松坤; 马睿和; 王春鹏; 李健; 陈锡蓉
Original assignee: Qilu University of Technology
Current assignee: Qilu University of Technology
Priority date: 2023-02-24
Filing date: 2023-02-24
Publication date: 2023-06-23

Abstract

The present disclosure provides a two-dimensional reversible information hiding method and system for images based on double ordering, which relates to the technical field of communication encryption, and the quantized DCT coefficients are obtained by obtaining an original carrier JPEG image for decoding; selecting embedded image blocks according to a global ordering method in double-layer ordering, and determining an information embedding order; calculating the gradient direction of the target block by using a local sequencing method, and calculating the predicted value of the target block by using the gradient direction; selecting an embedded coefficient, selecting a prediction error coefficient point to embed secret information through coefficient preprocessing and an embedding algorithm, and performing entropy coding on the carrier DCT coefficient to generate a JPEG image carrying secret information; when the secret data is extracted, the inverse application of the two-dimensional mapping chart is adopted, so that the extraction of the secret information and the lossless recovery of the image are realized. The method disclosed by the invention improves the visual quality of the image and reduces the size of the file.

Description

Image two-dimensional reversible information hiding method and system based on double ordering

Technical Field

The disclosure relates to the technical field of communication encryption, in particular to an image two-dimensional reversible information hiding method and system based on double ordering.

Background

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

With the popularization of computer networks, more and more files need to be transmitted through the networks, and data cannot be stared by some lawbreakers in the network transmission process, and the transmitted data can face a series of potential safety hazards such as tampering, stealing, infringement, illegal theft, counterfeiting, attack and the like. To solve this problem, secret information to be transmitted is often hidden in some carrier images, thereby ensuring the security of the information to be transmitted. In some related fields such as military, medical and digital evidence obtaining, which have high requirements on the integrity of the carrier image, the original carrier can be restored without damage after the secret information is extracted by the receiver. The reversible information hiding algorithm is an important branch in the information hiding field, can ensure that secret information is extracted and simultaneously accurately and undistorted recover carrier images, and is widely focused by people and becomes a research hotspot in the information security research field.

In recent years, with the rapid development of digital imaging and image compression techniques, JPEG has become the most widely used image recording format. The JPEG compression standard becomes the standard storage and transmission format of the current digital imaging equipment such as mobile phones, cameras and the like, so that the research on the reversible information hiding algorithm for JPEG images has higher application value and actual requirements. Considering the file format characteristics of compressed images, some traditional reversible information hiding algorithms are difficult to be directly applied to JPEG images, and according to the coding principle of the JPEG images, the reversible information hiding algorithms of the JPEG images are mainly divided into three types: reversible information hiding scheme based on modified quantization table, reversible information hiding scheme based on modified Huffman table, reversible information hiding scheme based on modified quantized DCT coefficient. Fridrich et al first propose an RDH algorithm based on modifying a quantization table, which embeds a bit of secret information by modifying a DCT coefficient losslessly. Chang et al modify specific elements in the quantization table to embed secret information using intermediate frequency coefficients of the quantization table. Wang divides some values in the original quantization table by a certain integer to shrink, and multiplies the corresponding quantized DCT coefficient by the same integer to expand, while the visual quality and embedding capacity are improved, the file increment after embedding the secret information becomes very large. Mobaseri et al propose a reversible information hiding algorithm based on modifying a Huffman table, and directly modify a part of the entropy-decoded JPEG image to embed secret information; wu and Deng propose a new integer vector conversion algorithm, and apply it to the JPEG huffman table to realize the synchronous change of bit stream and embedded message, such reversible information hiding algorithm based on modifying the JPEG image of huffman table can ensure smaller JPEG file increment and better image quality, but the embedded capacity is difficult to be improved.

The inventors found that the prior art also has the following problems:

considering the file format characteristics of the compressed image, some traditional reversible information hiding algorithms are difficult to directly apply to the JPEG image, and the file increment brought by embedding secret information is very large, or on the premise of ensuring the file increment, but the embedded capacity is difficult to promote.

Disclosure of Invention

In order to solve the problems, the present disclosure provides a two-dimensional reversible information hiding method and system for an image based on double ordering, wherein the method comprises the steps of analyzing DCT coefficient distribution, constructing a DCT coefficient prediction model by utilizing correlation of coefficients at the same position of adjacent DCT coefficient blocks, calculating gradient direction of a current image block, calculating a predicted value of the image block by utilizing the gradient direction of the predicted block, and realizing secret information extraction and lossless recovery of the image.

According to some embodiments, the present disclosure employs the following technical solutions:

the image two-dimensional reversible information hiding method based on double ordering comprises the following steps:

acquiring an original carrier JPEG image, and decoding to acquire quantized DCT coefficients;

selecting embedded image blocks according to a global ordering method in double-layer ordering, and determining an information embedding order;

calculating the gradient direction of the target block by using a local sequencing method, and calculating the predicted value of the target block by using the gradient direction; selecting an embedded coefficient, selecting a prediction error coefficient point to embed secret information through coefficient preprocessing and an embedding algorithm, and performing entropy coding on the carrier DCT coefficient to generate a JPEG image carrying secret information;

when the secret data is extracted, the inverse application of the two-dimensional mapping chart is adopted, so that the extraction of the secret information and the lossless recovery of the image are realized.

an image two-dimensional reversible information hiding system based on double ordering, comprising:

the image processing module is used for obtaining an original carrier JPEG image, decoding and obtaining quantized DCT coefficients;

the data embedding module is used for selecting embedded image blocks according to a global ordering method in double-layer ordering and determining an information embedding sequence; calculating the gradient direction of the target block by using a local sequencing method, and calculating the predicted value of the target block by using the gradient direction; selecting an embedded coefficient, selecting a prediction error coefficient point to embed secret information through coefficient preprocessing and an embedding algorithm, and performing entropy coding on the carrier DCT coefficient to generate a JPEG image carrying secret information;

and the data extraction module is used for realizing the lossless recovery of the secret information extraction and the image by adopting the inverse application of the two-dimensional mapping graph when the secret data is extracted.

a computer readable storage medium having stored therein a plurality of instructions adapted to be loaded by a processor of a terminal device and to perform the dual ordering based image two-dimensional reversible information hiding method.

a terminal device comprising a processor and a computer readable storage medium, the processor configured to implement instructions; the computer readable storage medium is for storing a plurality of instructions adapted to be loaded by a processor and to perform the dual ordering based image two-dimensional reversible information hiding method.

Compared with the prior art, the beneficial effects of the present disclosure are:

the method for dividing the prediction error into areas and further combining and embedding more 0 and 1 peak points into secret information is provided, more prediction error coefficient pairs capable of being embedded into the secret information can be generated in the area dividing mode, the zig-zag scanning is utilized, the positions of the last non-zero value are ordered, and the secret information is embedded into the flat sliding block as much as possible, so that unnecessary displacement caused by the embedding of the secret information can be reduced. In the information embedding stage, a two-dimensional histogram embedding scheme is adopted in the information embedding method, so that the information embedding capacity is improved.

According to the method, quality factors of images are respectively compressed, particularly when the embedding capacity is smaller, the improving effect of the image quality is more obvious, from the perspective of DCT coefficient distribution, an inter-block prediction model based on DCT coefficient distribution characteristics is built, the prediction precision is improved, a double-layer sequencing model is built for DC coefficients, and the average value taking operation is performed in the current gradient direction, so that the visual quality of the images is improved, and meanwhile, the size of a file is reduced.

In order to reduce image distortion and improve the embedding capacity of images, the two-dimensional embedding is applied to the JPEG domain, so that the problem of pixel overflow caused by embedding information in a space domain is avoided. And in the information embedding process, carrying out translation operation on the prediction error coefficient pair according to a specific two-dimensional mapping diagram, and embedding secret information into the carrier image. By applying two-dimensional embedding in the JPEG domain, the embedding capacity and visual quality of images are further improved compared to the one-dimensional embedding method.

Drawings

The accompanying drawings, which are included to provide a further understanding of the disclosure, illustrate and explain the exemplary embodiments of the disclosure and together with the description serve to explain the disclosure, and do not constitute an undue limitation on the disclosure.

FIG. 1 is a flow chart of secret information embedding in an embodiment of the present disclosure;

fig. 2 is a schematic diagram of a JPEG compression flow in an embodiment of the disclosure.

FIG. 3 is a luminance quantization table and a chrominance quantization table of a conventional JPEG;

FIG. 4 is a diagram of DCT block coefficients for an embodiment of the present disclosure;

fig. 5 is a DCT coefficient map (qf=70) of an image Lena of an embodiment of the disclosure;

FIG. 6 is a diagram of an example non-zero coefficient base point selection for a DCT block according to an embodiment of the disclosure;

FIG. 7 is a schematic diagram of gradient prediction according to an embodiment of the present disclosure;

FIG. 8 is a one-dimensional prediction error histogram of different images according to an embodiment of the present disclosure;

FIG. 9 is a partial two-dimensional prediction error histogram (0, 1) of an embodiment of the present disclosure;

FIG. 10 is a partial two-dimensional prediction error histogram (0, -1) of an embodiment of the present disclosure;

FIG. 11 is a two-dimensional embedding map of an embodiment of the present disclosure.

The specific embodiment is as follows:

the disclosure is further described below with reference to the drawings and examples.

It should be noted that the following detailed description is illustrative and is intended to provide further explanation of the present disclosure. Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this disclosure belongs.

It is noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of exemplary embodiments in accordance with the present disclosure. As used herein, the singular is also intended to include the plural unless the context clearly indicates otherwise, and furthermore, it is to be understood that the terms "comprises" and/or "comprising" when used in this specification are taken to specify the presence of stated features, steps, operations, devices, components, and/or combinations thereof.

Example 1

In one embodiment of the present disclosure, there is provided a two-dimensional reversible information hiding method for an image based on double ordering, including the steps of:

step one: acquiring an original carrier JPEG image, and decoding to acquire quantized DCT coefficients;

step two: selecting embedded image blocks according to a global ordering method in double-layer ordering, and determining an information embedding order;

step three: calculating the gradient direction of the target block by using a local sequencing method, and calculating the predicted value of the target block by using the gradient direction; selecting an embedded coefficient, selecting a prediction error coefficient point to embed secret information through coefficient preprocessing and an embedding algorithm, and performing entropy coding on the carrier DCT coefficient to generate a JPEG image carrying secret information;

step four: when the secret data is extracted, the inverse application of the two-dimensional mapping chart is adopted, so that the extraction of the secret information and the lossless recovery of the image are realized.

In step one, the original carrier JPEG image is obtained and then preprocessed, first, a color space conversion is performed to convert the image represented by the RGB space model into the image represented by the YCrCb space model, and then sampling is performed according to a set data sampling ratio.

Specifically, the color space is converted, the colors can be represented by independent attributes, each attribute is regarded as a coordinate base to form a color space, each color can be represented by a point in the space, and generally three independent attributes can completely represent one color space; taking a common RGB color space as an example, the color space takes three basic colors of R (Red), G (Green) and B (Blue) as coordinate bases, and other colors are mixed by utilizing the three coordinate bases in different degrees, so the color space is commonly called a three-primary-color mode.

In the JPEG algorithm, an image represented by an RGB space model needs to be converted into a YCrCb space model (an original image is represented by luminance and chromaticity), which is a compression transmission model widely applied to pictures and videos mainly considering human perceptibility, and the image occupies less bandwidth by compressing chromaticity information. The conversion relationship between RGB and YCrCb is shown in formula (1):

wherein R is the pixel value of the red channel; g is the pixel value of the green channel; b is the pixel value of the blue channel; y is brightness, namely a gray scale value; cb is the difference between the blue part of the RGB input signal and the luminance value of the RGB signal.

Cr is a value reflecting the difference between the red portion of the RGB input signal and the luminance value of the RGB signal. In general, the C value (including Cb, cr) should be a signed number, but by adding 128 to it, it is changed to an unsigned integer of 8 bits, thereby facilitating the storage and calculation of data. The inverse transformation is shown in formula (2):

next, the above-mentioned sampling is performed according to the set data sampling ratio, and because the sensitivity of human eyes to brightness conversion is much higher than that to color change, two sampling modes are generally sampled in the sampling process: YUV411 and YUV422 represent the meaning of data sampling ratios of three components Y, cb, cr, respectively, typically 4:1:1 or 4:2:2. Such sampling methods, while losing some accuracy, reduce the amount of data stored in a range that is imperceptible to the human eye.

In the second step, the process of selecting the embedded image block according to the global ordering method in the double-layer ordering is as follows: the original image data is segmented before DCT conversion, the image is divided into 8 x 8 small blocks, and after the image is divided, each sub-block is independently encoded.

Specifically, in the image processing process of DCT transformation, the image block is operated, so that the correlation among image pixels can be fully utilized, and the processing scale can be reduced. Therefore, the original image data must be segmented before DCT transformation is performed. The specific operation is to divide the image into 8 x 8 small blocks, which can be filled with black borders if the length or width is not an integer multiple of 8. The reason for the 8 x 8 partition is chosen because 8 x 8 is the largest size that an integrated circuit can support when formulating the JPEG standard, and the 8 x 8 size works best. After image segmentation, an independent encoding process will be performed for each sub-block.

The DCT is a discrete cosine transform (DCT for Discrete Cosine Transform), which is a transformation method for transforming an image into a frequency domain commonly used in image processing, and has a certain relation with the Fourier transform, wherein the transformation selected by the Fourier transform is a sine function and a cosine function, the transformation base selected by the cosine function transform is only a cosine function, and the two-dimensional discrete cosine transform is calculated by a formula (3):

where i, j is the position in the 8 x 8 matrix before transformation; u, v are the corresponding positions in the transformed 8 x 8 matrix; f (i, j) is a pixel value at a position of an i-th row and a j-th column; f (u, v) is a value for outputting the up-conversion result at the position of the u th row and the v th column; wherein,,

the corresponding inverse transform is:

where i, j, u, v=0, 1, …,7 indicate positions in the 8×8 matrix, and F (i, j) and F (u, v) correspond to magnitudes at the respective positions.

And then carrying out quantization processing, wherein in order to achieve the aim of removing irrelevant information, the DCT coefficient is required to be quantized, in order to ensure the visual quality of the image, the low-frequency part of the DCT coefficient is quantized by using a quantization step with a small value, the high-frequency part is quantized by using a quantization step with a large value, and a large amount of 0 can appear in the high-frequency part after quantization. DCT coefficient F obtained after quantization ^Q The calculation process of (u, v) is as shown in formula (5):

q (u, v) is the quantization step size at the coordinates (u, v), which is an element in the quantization table;

in order to correspond to the block size of the image and 64 DCT coefficients in the block one by one with JPEG compression, the quantization table size is also 8×8, and the values of the elements in the table are different due to different positions, so as to ensure the compression degrees of the coefficients in different frequency regions to different degrees.

Meanwhile, because the sensitivity of human eyes to brightness and chromaticity is different, the compression algorithm also has quantization tables which are respectively different for brightness and chromaticity, and respectively correspond to different compression degrees. Specifically, the quantization step size value in the luminance quantization table is smaller, and the quantization step size value in the chrominance quantization table is larger. Therefore, the quantization phase requires two different 8 x 8 quantization matrices for processing luminance information and for processing chrominance information, respectively. Fig. 3 shows a luminance quantization table and a chrominance quantization table of a conventional JPEG, and in practical applications, a coefficient is often multiplied on the basis of a quantization coefficient, and the size of the coefficient is adjusted to obtain different compression ratios.

The JPEG standard employs a zig-zag scanning strategy.

Step three, calculating the gradient direction of the target block by using a local sequencing method, and calculating the predicted value of the target block by using the gradient direction; selecting an embedded coefficient, selecting a prediction error coefficient point to embed secret information through coefficient preprocessing and an embedding algorithm, and performing entropy coding on the carrier DCT coefficient to generate a JPEG image carrying secret information;

firstly, constructing a double-layer ordering model by using DC coefficients, wherein the first layer is global ordering in double-layer ordering, and the global ordering method in double-layer ordering is to use the number of non-zero AC coefficients of a current DCT block as smoothness measurement of the current block, and preferentially embed DCT blocks with fewer non-zero coefficients.

The second layer is local sorting, and the local sorting method is to sort adjacent blocks of the target block by using the DC coefficient of each block, and preferably select DCT blocks with higher correlation as optimal prediction blocks.

Specifically, the prediction error algorithm based on DCT coefficient multi-stage ordering provided by the disclosure analyzes the distribution characteristics of the DCT coefficients of the JPEG image, and constructs an inter-block prediction model according to the characteristic of larger inter-block correlation of DCT. In order to improve the algorithm performance, an optimization strategy is designed, and a DC coefficient is utilized to construct a double-layer ordering model, namely a global ordering scheme and a local ordering scheme. The global ordering uses the number of non-zero AC coefficients of the current DCT block as the smoothness measurement of the current block, and the DCT blocks with fewer non-zero coefficients are preferentially embedded. The local ordering is to order the adjacent blocks of the target block by using the DC coefficient of each block, and the DCT block with higher correlation is preferentially selected as the optimal prediction block, so as to reduce some ineffective shift.

There is a correlation between the DCT coefficients in each block, whether the distribution of the DCT coefficients in a JPEG image is within a single block (intra-block correlation) or between two blocks (inter-block correlation). Wherein within each 8 x 8 block of DCT coefficients, the magnitude of the AC coefficients decreases with increasing frequency, which indicates that there is a certain correlation between the magnitudes of adjacent DCT coefficients within the same DCT block. By Zigzag scanning the coefficients of each 8 x 8 sub-block, it is not difficult to find that most of the high frequency AC coefficient values are zero, as in fig. 4, and it follows that the correlation within the DCT block decreases with increasing frequency. In comparison to intra-block correlations, the coefficients in the current DCT block remain correlated with the coefficients of their neighboring blocks, commonly referred to as local intra-block (or inter-subband) correlations or inter-block correlations. The coefficient value on a certain DCT block has a significant correlation with the coefficient value at the same position as that of its neighboring block, as shown in fig. 5, the inter-block correlation is generally slightly stronger than the inter-block correlation, but as the frequency position in the current DCT block increases and the distance from the target block increases, the inter-block correlation also gradually decreases, so that in the process of predicting the DCT coefficients, the inter-block correlation using good DCT coefficients is important.

Equation (6) may calculate the correlation between neighboring blocks. Wherein X is ^m And X ⁿ Each representing two adjacent blocks of DCT coefficients,

and->

Representing coefficients in a block->

And->

Respectively represent the average value of adjacent block coefficients, and the final formula yields the value Corr (X ^m ,X ⁿ ) The magnitude of the correlation between two blocks can be measured.

m and n respectively represent different adjacent blocks;

in the data embedding process, in order to improve the prediction performance of the predictor, a double-layer sequencing technology is adopted, and after double-layer sequencing, the predictor can generate more embeddable prediction error values, so that the distortion caused by invalid displacement is reduced, and the embedding capacity is increased.

The double-layer ordering model is divided into two layers: the first layer is global sorting, and the embedded blocks are sorted according to the number and the correlation of non-zero coefficients in the image; the second layer is local ordering, gradient prediction is carried out on the target block by utilizing DCT blocks around the target block, and the optimal prediction block of the target block is selected for prediction. The double layer ordering is described in detail below.

The main purpose of global ordering is to find a flat slider of the image and to perform embedding preferentially. As can be seen from the observation of fig. 5, in the smooth region of the image, the absolute values of most non-zero coefficients are very small, so that more zero values and smaller prediction error values can be obtained after coefficient prediction, and more prediction error values are distributed near the zero values; in the complex texture region of the image, the absolute value of the non-zero coefficient is larger, the AC coefficient is predicted, and the obtained predicted value is more scattered. Aiming at the characteristics, the algorithm takes the number of non-zero coefficients in each block as a smoothness measurement standard of each block, so that smoothness ordering is carried out on each block, and flat sliding blocks are preferably selected for embedding in the information embedding process. The algorithm keeps the last non-zero coefficient in each block unchanged during the embedding process by selecting the point as a reference point in order to ensure the reversibility of the algorithm. And performing zigzag scanning on each DCT block, changing the quantized DCT coefficient from two dimensions to one dimension through scanning, setting the last non-zero coefficient in the one-dimension data as a reference point, and setting the position of the last non-zero coefficient as a threshold value T. For ease of understanding, fig. 6 gives an example of reference point selection for three DCT blocks.

As can be seen from fig. 6, the last non-zero coefficient values of the three DCT blocks are respectively: "7, -1, 1", and the DCT block is zigzagged, the two-dimensional data becomes one-dimensional data, so that the threshold T is "1, 7, 21". And (3) globally sequencing the whole image according to the threshold value T of each block, sequencing DCT blocks from small to large according to the threshold value T, and preferentially embedding the blocks with smaller T values. Meanwhile, in order to keep the algorithm reversible in the embedding process, the DCT coefficient corresponding to the T value is not changed.

Wherein the distortion caused by the embedding of data is related not only to the DCT coefficients within the block but also to the correlation of neighboring blocks. Thus, if there are two or more DCT blocks with the same threshold, the DCT blocks are secondarily ordered by calculating the correlation of the current block with neighboring blocks. Firstly, 8 DCT blocks in four directions of horizontal, vertical, 45 degrees and 135 degrees are selected, the correlation is calculated by utilizing the difference value of the current DC coefficient and the DC coefficient of the adjacent block, and the formula for calculating the correlation is as (7):

wherein C is _i,j Mean value representing correlation with neighboring blocks, D _i-m,j-n Representing the DC coefficients of eight adjacent blocks up, down, left, and right. If C _i,j The larger the value, the smaller the correlation between the block and the adjacent block, and conversely, the larger the correlation.

There is a correlation between neighboring blocks in a JPEG image, but the correlation of neighboring blocks in different directions is different. In order to make the predicted coefficient value more accurate, the optimal gradient direction is selected by utilizing the gradient change trend of the DC coefficient. And counting and comparing the DC coefficients, respectively calculating the gradient trend of the block in the horizontal direction and the gradient trend in the vertical direction, and predicting the current block according to the gradient trend, so that the accuracy of prediction is improved.

The specific method comprises the following steps: taking fig. 7 as an example, assume that the current block is x _i,j First, the adjacent block x of the current block and the horizontal direction is calculated _i,j-1 ，x _i,j+1 And adjacent blocks x in the vertical direction _i-1,j ，x _i+1,j Gradient D of (2) ₁ And D ₂ The formula is as follows:

D＝min(D ₁ ,D ₂ ) (10)

the symbols in the formulas (8) and (9) correspond to fig. 7, and D is the direction of the smallest gradient from the current block selected last. After the gradient direction of the current block is selected, the AC coefficients of adjacent blocks in the same gradient direction of the current block are subjected to average operation, and the obtained value is the predicted value of the AC coefficient of the current block, and the formula is as follows:

wherein,,

is the predicted value of the DCT coefficient. In this algorithm, the DC coefficients of all DCT blocks remain unchanged during the embedding process. The receiver can therefore keep the DC-coefficient unchanged during the process of extracting the secret information and recovering the carrier image.

As an embodiment, the process of selecting the prediction error coefficient points to embed the secret information is to firstly perform regional processing on the prediction errors, perform pairwise grouping on the prediction error values in the same region to form two-dimensional errors, namely divide the prediction errors with the pre-error values of 0 and 1 into inner regions, perform pairwise grouping to form a two-dimensional prediction error histogram, divide other prediction error values except the inner regions into outer regions, perform pairwise pairing, and make room for embedding the secret information by translating the prediction error pairs of the outer regions.

Because, fig. 8 shows a one-dimensional prediction error histogram of which the quality factors of the image Lena, baboon, barbara, peppers, bridge, splash obtained by the prediction error algorithm are 70, and the local area map with the error prediction values of-5 to 5 is enlarged, it can be easily seen from fig. 8 that the one-dimensional prediction error histogram has peak points at "0" and also has higher prediction error distribution at "±1". Tables (1), (2) and (3) are local error prediction tables of 24 images of the Kodak library with quality factors of 70, 80 and 90 respectively, wherein the peak points of the one-dimensional prediction error histogram with red values in the tables are also the number of the 0 points with the prediction error value, and the distribution ratio is higher when the blue value points are the one-dimensional prediction error value equal to +/-1. Inspired by the adaptive pixel pairing strategy proposed by Ou et al, the present disclosure proposes a new prediction error pairing strategy according to the distribution of the one-dimensional prediction error histogram: the prediction errors are divided into regions according to the magnitude of the values, the prediction errors having values equal to "0, 1" or "0, -1" are combined to form an inner region, and the prediction error values of the other remaining cases are combined to form an outer region.

After the prediction errors are subjected to the region division processing, two pairs of prediction error values in the same region are subjected to two-by-two pairs to form a two-dimensional prediction error, the quality factors of the image Lena, baboon, barbara, peppers, bridge, splash are respectively 70, the two-dimensional prediction error histogram formed by two pairs of the internal regions is formed by dividing the ' 0 ', the ' 1 ' and the ' 0 ' -1 ', wherein red numbers represent peak points of the two-dimensional prediction error histogram formed by the internal regions, the two-by-two pairs of the two-dimensional prediction error histogram formed by dividing the ' 0 ' -1 ' into the internal regions are easily found by comparing the two-by-two images of the figure 9 and the figure 10, the peak points formed by two-by-two pairs of the two-dimensional prediction error histogram formed by dividing the ' 0 ' -1 ' into the internal regions are higher than the peak points formed by two-by two pairs of the prediction error values of the ' 0 ' -1 ', and the blue numbers formed by two-by two pairs of the two-by dividing the ' 0 ' -1 ' into the two-dimensional prediction error values of the figure 10 are smaller than that by another method, so that the image distortion values caused by embedding secret information into the peak point positions are smaller, and the invalid displacement is reduced by comparison. Other prediction error values except the inner region are divided into the outer region, two pairs are performed, and space is made for secret information embedding by translating the prediction error pairs of the outer region.

The secret information is embedded into the (0, 0) peak point of the two-dimensional prediction error histogram, the two-dimensional mapping is improved, and the algorithm divides the prediction error difference with the value of 0 and minus 1 into an internal area, and the secret information is embedded into the two-dimensional prediction error histogram peak point formed by the internal area, so that the conventional two-dimensional mapping is correspondingly changed, the movement rules of the other quadrants are deleted, and only the third quadrant is reserved. Fig. 11 is a modified two-dimensional map table in which coordinate points are divided into two types, a type a point for embedding secret information, and B type point for panning, making room for secret information embedding.

It is assumed that the embedded secret information is a binary sequence (other forms of secret information may be converted into binary sequences).

Type a: the type a is a red dot in the two-dimensional map, wherein the red dot can embed one or two bits of secret information according to the difference of secret information, the dot remains unchanged if the secret information to be embedded is "0", the next secret information is scanned if the secret information to be embedded is "1", the dot is horizontally shifted by one bit to become "(-1, 0)", and the dot is vertically shifted by one bit to become "(0, -1)", if the secret information of the next bit is "1").

Type B: the type B dots are blue dots in the two-dimensional map, and as can be seen by viewing the blue dots in fig. 11 make room for embedding secret information by horizontal or vertical translation.

Type C: the C-type point is a white point in the two-dimensional map and is not moved in any way in order to reduce unnecessary distortion due to translation.

When the secret data is extracted, the steps of extracting the secret information and recovering the image in a lossless manner are realized by adopting the inverse application of the two-dimensional mapping chart, wherein the steps are as follows:

step 1: entropy decoding the secret-carrying JPEG image to obtain a secret-carrying DCT coefficient thereof;

step 2: calculating the gradient direction of the target block by using a global ordering method and a local ordering method, and calculating a prediction error value;

step 3: dividing the prediction error value into areas and matching coefficients according to a coefficient matching strategy;

step 4: extracting secret information by using an inverse application of the two-dimensional map, and recovering the AC coefficients;

step 5: after the secret information is extracted, entropy coding is carried out on the DCT coefficient, and finally the original JPEG image with lossless recovery is obtained.

Example 2

In one embodiment of the present disclosure, there is provided a two-dimensional reversible information hiding system for images based on double ordering, comprising:

Example 3

In one embodiment of the present disclosure, a computer readable storage medium is provided, in which a plurality of instructions are stored, said instructions being adapted to be loaded by a processor of a terminal device and to perform said one image two-dimensional reversible information hiding method step based on double ordering.

Example 4

In one embodiment of the disclosure, a terminal device is provided, including a processor and a computer readable storage medium, where the processor is configured to implement instructions; the computer readable storage medium is for storing a plurality of instructions adapted to be loaded by a processor and to perform the steps of the method for image two-dimensional reversible information hiding based on double ordering.

The above examples 2, 3, 4 specifically perform the method steps of example 1.

The present disclosure is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the disclosure. It will be understood that each flow and/or block of the flowchart illustrations and/or block diagrams, and combinations of flows and/or blocks in the flowchart illustrations and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.

These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.

While the specific embodiments of the present disclosure have been described above with reference to the drawings, it should be understood that the present disclosure is not limited to the embodiments, and that various modifications and changes can be made by one skilled in the art without inventive effort on the basis of the technical solutions of the present disclosure while remaining within the scope of the present disclosure.

Claims

1. The image two-dimensional reversible information hiding method based on double ordering is characterized by comprising the following steps of:

2. The two-dimensional reversible information hiding method based on double ordering according to claim 1, wherein the JPEG image is preprocessed before the original carrier JPEG image is obtained for decoding, the image represented by the RGB space model is converted into the image represented by the YCrCb space model by first performing color space conversion, and then sampling is performed according to a set data sampling ratio.

3. The two-dimensional reversible information hiding method of image based on double ordering according to claim 1, wherein the process of selecting embedded image blocks according to global ordering method in double ordering is: the original image data is segmented before DCT conversion, the image is divided into 8 x 8 small blocks, and after the image is divided, each sub-block is independently encoded.

4. The two-dimensional reversible information hiding method of image based on double ordering according to claim 1, wherein a double-layer ordering model is constructed by using DC coefficients, the first layer is global ordering in double-layer ordering, and the method of global ordering in double-layer ordering is to use the number of non-zero AC coefficients of the current DCT block as smoothness measure of the current block, and preferentially embed DCT blocks with few non-zero coefficients.

5. The method for two-dimensional reversible information hiding of image based on double ordering according to claim 4, wherein the second layer is local ordering, and the method of local ordering uses DC coefficient of each block to order adjacent blocks of target block, and preferably selects DCT block with higher correlation as optimal prediction block.

6. The method for hiding two-dimensional reversible information of image based on double sequencing as claimed in claim 1, wherein the process of selecting prediction error coefficient points to embed secret information is to firstly process prediction errors in regions, to process two-by-two groups of prediction error values in the same region to form two-dimensional errors, namely, divide the prediction errors with the pre-error values of 0, -1 into inner regions, and process two-by-two groups of groups to form a two-dimensional prediction error histogram, divide other prediction error values except the inner regions into outer regions, process two-by-two pairs, and make room for embedding secret information by translating the prediction error pairs of the outer regions.

7. The method for hiding two-dimensional reversible information of image based on double ordering according to claim 1, wherein when secret data is extracted, the steps of implementing the extraction of secret information and lossless recovery of image by using inverse application of two-dimensional map are as follows:

8. An image two-dimensional reversible information hiding system based on double ordering, characterized by comprising:

9. A computer readable storage medium, characterized in that a plurality of instructions are stored, which instructions are adapted to be loaded by a processor of a terminal device and to perform the dual ordering based image two-dimensional reversible information hiding method as claimed in any one of claims 1-7.

10. A terminal device comprising a processor and a computer readable storage medium, the processor configured to implement instructions; a computer readable storage medium for storing a plurality of instructions adapted to be loaded by a processor and to perform the dual ordering based image two-dimensional reversible information hiding method as claimed in any one of claims 1 to 7.