CN113421181A - Information hiding method based on pre-estimated robustness - Google Patents

Abstract

The invention discloses an information hiding method based on pre-estimated robustness, which comprises the following steps: setting the coefficient set C to { C in units of each k × k image block_f，f＝1，2，…，3k²(v) 4) as candidate vector; and calculating the estimated robustness of each candidate carrier according to a robustness estimation function r (m, f, t), and selecting the carrier with the optimal robustness to carry secret information. Wherein the robustness pre-estimating function r (m, f, t) comprises the extraction robustness r brought by the embedded modification_mExtraction robustness r brought by spatial frequency of embedded carrier_fExtraction robustness r due to texture characteristics_t. Compared with the prior art, the method has the advantages that the information is embedded in a Discrete Cosine Transform (DCT) domain of the block image, and the strong robust region is selected for embedding by analyzing the texture property and the frequency characteristic of the carrier and the influence of embedding modification on the robustness of the information extraction after embedding, so that the robustness of information hiding is improved.

Description

Information hiding method based on pre-estimated robustness

Technical Field

The invention relates to the technical field of digital image information hiding, in particular to an information hiding method based on pre-estimated robustness.

Background

Due to the rapid development of the internet, the wide use of image data, and the emergence of various powerful image processing software in recent years, the processes of digital image transmission, storage, copying, editing, etc. become convenient and fast, and people also begin to pay attention to the multimedia information security problem. The information hiding technology is an information security technology for hiding secret information in an existing medium, and the more redundant information of the medium, the more easily the secret information is hidden, the more security is. The history of information hiding can be traced back to steganography 440 years ago, and in the digital era, the information hiding technology has become a hotspot in the research field of information security.

The image information hiding method is divided into methods of histogram translation, spectrum expansion, bit plane substitution and the like according to the embedding position, and the information extraction robustness is an important performance index no matter which information embedding method is adopted. At present, robustness is usually measured by comparing the accuracy of extracted information after the information is extracted by a receiving end.

The document "On the fault-complete performance for a class of robust image robust graphics" (. Zhang, c.qin, w.m.zhang, f.l.liu, and x.y.luo, Signal Processing, 146: 99-111, 2018) proposes an error model based On burst errors and STC decoding errors to analyze the image error tolerance performance, and also describes the error of the encrypted image received by the receiving end and the steganography robust performance. The document "firmware distribution image reconstruction" (j.tao, s.li, x.zhang, and z.wang. ieee transformations on Circuits and Systems for Video Technology, 29 (2): 594-600, 2019.) proposes to modify the carrier image by adjusting the image compression factor to achieve embedding of information, so that the dense image is the same as the image after true compression, thereby improving robustness.

Disclosure of Invention

The invention aims to provide an information hiding method based on pre-estimated robustness, which can select a strong robust region for embedding by measuring the influence of the texture and the spatial correlation of the region where an embedded carrier is located on the extracted robustness, thereby improving the robustness of information hiding.

The technical scheme for realizing the purpose of the invention is as follows:

an information hiding method based on pre-estimated robustness takes each k multiplied by k image block as a unit and sets a coefficient set C to { C ═ C_f，f＝1，2，…，3k²(v) 4) as candidate vector; wherein, C_fIs a DCT coefficient with spatial frequency f; k is a natural number and k is more than 1;

calculating the estimated robustness of each candidate carrier according to a robustness estimation function r (m, f, t), and selecting the carrier with the optimal robustness to carry secret information, namely

f^*＝arg max r(m，f，t)subjectto：f∈{1，2，…，3k²/4}；

The robustness pre-estimation function

r(m，f，t)＝r_m+r_f+r_t；

Wherein the content of the first and second substances,

r_m＝αe^mthe extraction robustness brought about by the embedding modification, m represents the embedding modifier,

C_f' embedding the modified carrier coefficients; alpha represents a weight coefficient for leading the embedding modification to be robust;

r_f＝βe^1-fextracting robustness brought by the spatial frequency of the embedded carrier, wherein f represents the 1-dimensional frequency of the DCT domain of the image block, and beta represents a weight coefficient of the robustness brought by the spatial frequency;

the extraction robustness brought by the texture characteristics is achieved, t represents the texture characteristics of the image block where the coefficient is located, the values are 0, 1 and 2 respectively represent flat, edge and texture areas, and gamma represents a weight coefficient brought by the texture characteristics and the robustness.

Compared with the prior art, the invention has the advantages that,

information is embedded in a Discrete Cosine Transform (DCT) domain of a block image, and a strong robust region is selected for embedding by analyzing the influence of texture attributes, frequency characteristics and embedding modification of a carrier on the extraction robustness of the embedded information, so that the robustness of information hiding is improved.

Detailed Description

The invention designs an evaluation function of the information extraction robustness of the DCT coefficient before information embedding by using the texture of the image block, the spatial frequency of the DCT coefficient in the block and the embedding modifier, and optimizes the robustness of the information embedding based on the robust function.

The invention includes a method for measuring robustness of information embedding.

The method mainly comprises the following steps:

firstly, on the basis of the influence change of an image block caused by information embedding, the shielding effect of carrier texture related to the information embedding on embedding modification is researched, and in an area with rich texture, the larger embedding modification cannot generate obvious distortion and can generate stronger robustness; secondly, the problem of modifying the corresponding amplitude and the pixel value in the range of the corresponding space domain image block is generated based on the modification of different space frequencies, the high-frequency coefficient corresponds to the high-frequency part of the space domain, the low-frequency coefficient corresponds to the low-frequency part of the space domain, and the high-frequency coefficient and the low-frequency coefficient of different textures have different influences, so that the optimal embedded carrier is selected from the low-frequency coefficient to improve the extraction robustness based on the principle that the low-frequency part is not easily modified; then, based on the influence of the embedded modifier on the extraction robustness, large embedded modification can better resist transmission errors, the modification rate of the modifier to the original coefficient is defined as the reference of the robustness, the balance between distortion and the robustness caused by embedding is searched, and the extraction robustness is maximized on the premise of not causing obvious distortion. And finally, based on the three factors, establishing a function relationship between the pre-evaluated robustness and the textuality, the spatial frequency and the modification quantity by using a function.

The method comprises the following specific steps:

suppose that the distortion cost of a particular element of a coefficient in an (i, j) th image block given a given image is evaluated.

1. Pre-estimated robust function

(1) Texture factor

Dividing an image into image blocks of k multiplied by k (k is more than 1), performing DCT (discrete cosine transformation), calculating the variance of AC (Alternating Current) coefficients of each block, and sequencing the blocks from large to small to form a block sequence { b }₁，b₂，…，b_nWhere n is the number of blocks in the image, the texture value of the ith block in the sequence of ordered blocks defines:

because the texture characteristic of the image has a direct relation with the spatial frequency coefficient, and meanwhile, the texture can provide better masking for embedding distortion, the texture rich area can provide stronger robustness, and texture factors are expressed as follows:

where gamma is the scaling factor.

(2) Spatial frequency factor

And after the DCT coefficients in the image blocks are scanned in a zigzag manner, 1DDCT coefficients are formed. The corresponding index of each DCT coefficient represents the spatial frequency within its corresponding block. The direct current coefficient reflects the average energy of the image block, the more the information embedding therein can resist the local information loss caused by information transmission, the information embedding has the strongest robustness, the robustness gradually decreases with the increase of frequency, and the trend of exponential decrease is presented, so the spatial frequency factor can be expressed as follows:

r_f＝βe^1-f

where β is the scaling factor.

(3) Embedding modifiers

The robustness of information extraction in image carriers is a countermeasure between modification by information embedding and transmission channel interference. If the information embedding modification quantity is larger than the interference quantity, the information can be correctly extracted, otherwise, error codes of the extracted information are caused. The amount of information embedding is therefore proportional to the extraction robustness, which can be expressed as:

r_m＝αe^m

where alpha is the scaling factor.

(4) Information extraction robust function

Based on the texture of the embedded carrier, the spatial frequency and the modification of the DCT coefficient, in combination with the property of the exponential function, after the information of the specified carrier is embedded, the receiving end information extraction robust function is defined as follows:

2. an information hiding algorithm based on pre-estimated robustness:

(1) candidate vectors

As mentioned above, information is embedded in k × k (k > 1) image blocks, only one DCT coefficient is selected on each image block as a carrier to carry information, and low-frequency coefficients in the image blocks subjected to DCT transformation have better robustness, so that the previous image blocks are selected

Low frequency coefficients as candidate vectors:

(2) selecting the optimal carrier in the block according to the robust function of information extraction

As mentioned above, the texture is determined by the variance of the AC coefficient of the block in the image, and takes values of 0, 1, and 2, which represent the flat, edge, and texture blocks, respectively; the spatial frequency of each DCT coefficient is the serial number after Z scanning corresponding to the DCT coefficient; because the values of different spatial frequency coefficients are different, the absolute value of the embedding modifier cannot reflect the influence generated by modification, and the embedding modification principle is defined as:

wherein C is_f' is embedding the modified carrier coefficients. Calculating the robustness of each candidate coefficient according to the information extraction robust function, and selecting the optimal coefficient embedding information:

(3) embedding information

And according to a specific embedding algorithm, embedding information on the selected optimal carrier.

Before embedding, the invention selects the optimal embedding carrier to carry information through the influence on distortion and robustness brought by embedding operation and the influence on the embedding operation caused by embedding modifier, texture characteristic and spatial frequency, thereby being beneficial to popularization and application of the invention.

The method has the advantages that the method can be verified by the quality of the secret video and the extraction accuracy rate under the noise environment, and mainly comprises the following steps:

1. extraction robustness under salt-pepper noise and Gaussian noise conditions

The information is embedded by a classical bit plane replacement method, and meanwhile, the test is carried out by a classical picture, and the test result is shown in table 1. Where "GN" refers to Gaussian Noise (Gaussian Noise) and "SP" refers to Salt and Salt Noise (Salt & Pepper).

Table 1: information extraction accuracy rate under noise interference

2. Effects of the invention

The results of comparing the extraction with and without the present invention after the interference of salt and pepper noise when using the same embedding method are shown in table 2. Wherein "N" and "Y" indicate the nonuse and use of the present invention, respectively.

Table 2: extraction accuracy before and after using the method under the interference of salt and pepper noise

Claims (1)

1. An information hiding method based on pre-estimated robustness is characterized in that,

setting the coefficient set C to { C in units of each k × k image block_f，f＝1，2，…，3k²(v) 4) as candidate vector; wherein, C_fIs a DCT coefficient with spatial frequency f; k is a natural number and k is more than 1;

calculating the estimated robustness of each candidate carrier according to a robustness estimation function r (m, f, t), and selecting the carrier with the optimal robustness to carry secret information, namely

f^*＝arg max r(m，f，t)subject to：f∈{1，2，...，3k²/4}；

The robustness pre-estimation function

r(m，f，t)＝r_m+r_f+r_t；

Wherein the content of the first and second substances,

r_m＝αe^mthe extraction robustness brought about by the embedding modification, m represents the embedding modifier,

C_f' embedding the modified carrier coefficients; alpha represents a weight coefficient for leading the embedding modification to be robust;

r_f＝βe^1-fextracting robustness brought by the spatial frequency of the embedded carrier, wherein f represents the 1-dimensional frequency of the DCT domain of the image block, and beta represents a weight coefficient of the robustness brought by the spatial frequency;

the extraction robustness brought by the texture characteristics is achieved, t represents the texture characteristics of the image block where the coefficient is located, the values are 0, 1 and 2 respectively represent flat, edge and texture areas, and gamma represents a weight coefficient brought by the texture characteristics and the robustness.