CN109543425B - Image data hiding method based on tensor decomposition - Google Patents
Image data hiding method based on tensor decomposition Download PDFInfo
- Publication number
- CN109543425B CN109543425B CN201811309996.9A CN201811309996A CN109543425B CN 109543425 B CN109543425 B CN 109543425B CN 201811309996 A CN201811309996 A CN 201811309996A CN 109543425 B CN109543425 B CN 109543425B
- Authority
- CN
- China
- Prior art keywords
- tensor
- singular
- order
- image
- decomposition
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F21/00—Security arrangements for protecting computers, components thereof, programs or data against unauthorised activity
- G06F21/60—Protecting data
- G06F21/602—Providing cryptographic facilities or services
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Software Systems (AREA)
- General Health & Medical Sciences (AREA)
- Computer Hardware Design (AREA)
- Computer Security & Cryptography (AREA)
- Health & Medical Sciences (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Bioethics (AREA)
- Image Processing (AREA)
- Editing Of Facsimile Originals (AREA)
Abstract
The invention discloses an image data hiding method based on tensor decomposition, which is characterized in that high-order singular value decomposition is applied to an image of required embedded information, the required embedded information replaces a left singular matrix element corresponding to a minimum singular value obtained by tensor decomposition, and an original image is reconstructed by the replaced singular matrix. The method comprises the following steps: putting an image of the required embedded data into a third-order cube tensor; decomposing the cube tensor by using a high-order singular value decomposition method to obtain a core tensor, singular values on each mode and a singular matrix; replacing a left singular matrix element corresponding to a minimum singular value under a certain mode obtained by tensor decomposition by the needed embedded information; reconstructing a third-order tensor by the three left singular matrixes and the core tensor which are embedded with the information; and extracting the image with the embedded information from the third-order tensor. The method applies tensor decomposition to image data hiding, improves the safety of information transmission, and simultaneously does not cause the reduction of image quality after extracting embedded information.
Description
Technical Field
The invention relates to a data hiding technology, in particular to an image data hiding method based on tensor decomposition.
Background
The data hiding technology is a comprehensive technology, which relates to the knowledge of multiple disciplines such as human physiology, computer graphics, cryptography, information processing, and the like, and mainly studies on how to hide confidential information in public information and then transmit the confidential information through the public information. The general data hiding technology changes original data in the process of information embedding, and damages the original data after extracting embedded information, so that the lossless data hiding technology becomes an important research direction of the data hiding technology.
The image data hiding technology is a branch of data hiding technology, and confidential information is embedded into a certain public image and is transmitted through the image in a mode of not attracting attention of attackers and being safer compared with a password technology. The existing image data hiding method comprises the following steps:
fridrich.j proposes a reversible information hiding method based on lossless data compression. The lossless data compression method mainly extracts and compresses LSB information of an original image, and a space generated after compression is used for embedding hidden information.
Tian proposes a reversible information hiding method based on difference value expansion. The basic idea of difference expansion is to hide information by expanding the difference between adjacent pixel points. The method proposed by Tian extends the difference value k 'by calculating' l =2k l +w l The method is easy to cause pixel value overflow, and the problem of pixel overflow can be solved by utilizing the positioning graph to realize image recovery, but the influence on the capacity of embedded information is large.
Ni proposes a reversible information hiding method based on histogram adjustment. The basic idea of histogram adjustment is to embed information using zero redundancy by adjusting the histogram of the original image. Ni proposes a method of embedding information using three maximum points and minimum points in an image histogram, which has a large capacity of embedded information but is highly complex, and the maximum points are not the best positions to embed information.
The invention provides an image data hiding method based on tensor decomposition, which is characterized in that an original RGB image is placed in a three-order cube tensor, the position of the three-order cube tensor is used as a key, the tensor is decomposed by using high-order singular value decomposition to obtain singular matrixes under three modes, a left singular vector corresponding to the minimum singular value in the mode determined by the size of the original image is used for replacing a column of elements with required embedded information, the three-order tensor is reconstructed by the core tensor obtained by decomposition and the left singular matrix under the three modes (including the left singular matrix of the embedded information in the mode), the RGB image after the information is embedded is obtained by using the position key, and the embedded information (the left singular matrix element corresponding to the minimum singular value in the mode) can be obtained by performing high-order singular value decomposition on the RGB image.
Disclosure of Invention
The invention aims to fully utilize the spatial structure in the RGB image, resist attack to a certain extent, increase the safety of embedded information and keep the image quality basically unchanged after the embedded information is extracted.
In order to achieve the above object, the present invention provides an image data hiding method based on tensor decomposition, which includes:
putting an original RGB image into a three-order cube tensor with random values, and taking the put position as a secret key;
decomposing the third-order tensor by using high-order singular value decomposition to obtain a core tensor and singular matrixes (comprising three left singular matrixes, three singular value matrixes and three right singular matrixes) under three modes respectively;
selecting a singular value matrix obtained by decomposition under a mode determined by the size of an original image to find a left singular matrix element corresponding to the minimum singular value in the singular value matrix, and replacing the row of elements with the required embedded information to obtain a new left singular matrix under the mode;
reconstructing a third-order tensor by using a left singular matrix and a core tensor under three modes, and extracting an image with embedded information through a key;
obtaining a singular matrix by decomposing the image of the embedded information through a high-order singular value, finding a left singular matrix element corresponding to a minimum singular value in the mode, wherein the column vector is the embedded information;
and after information is extracted, reconstructing the tensor by using the core tensor obtained by decomposition and the left singular matrix under the three modes, and finding the image after the information is extracted through the position key.
Wherein, according to the third order cube tensor, comprising:
the magnitude of the third order cube tensor is determined from the selected image: a = max (size (I)), I being the selected image;
the tensor element takes a value of 0-255;
the position of a third-order cube tensor is put into the image to serve as a key;
according to the method, the decomposition of the third-order cube tensor by the high-order singular value is carried out, and the method comprises the following steps:
decomposing the third-order cube tensor by a Tucker to obtain a core tensor and expansion matrixes under three modes;
SVD decomposition is carried out on the expansion matrixes under the three modes to obtain singular matrixes (a left singular matrix, a singular value matrix and a right singular matrix), and the matrixes corresponding to the three modes are the same in size.
Wherein reconstructing the third order tensor extracted image according to the embedded information comprises:
finding a left singular matrix column vector corresponding to the minimum singular value in the mode, and replacing the column vector with information to be embedded;
reconstructing a third-order tensor by the left singular matrix, the left singular matrix in other modes and the core tensor;
and extracting the original image embedded with the information through the position key.
Wherein, according to the extracting the embedded information, the method comprises the following steps:
putting the embedded information image into the original three-order cube tensor through a position key;
performing high-order singular value decomposition on the third-order tensor;
finding a left singular matrix element corresponding to the minimum singular value in the mode, wherein the column vector element is embedded information;
reconstructing a third-order tensor by the left singular matrix after the information is extracted, the left singular matrix in other modes and the core tensor;
and finding the image with the embedded information extracted in the third-order cube tensor through the position key.
Based on the invented method, compared with the existing method, the method has the outstanding differences and contributions that:
1. the invention relates to an image expression in a tensor form, which fully utilizes the spatial structure of an RGB image, the formed tensor obtains three modes through Tucker decomposition, more image details can be reserved in the next step of SVD decomposition by expanding the mode determined by the size of an original image, and the influence on the image quality after the embedded information is extracted is very small.
2. According to the three-order cube tensor related by the invention, the matrix size of the three modes obtained by Tucker decomposition after expansion is the same, when the three modes are attacked, an attacker cannot intuitively find out which mode is determined by the original image size, information embedded in a singular matrix under the mode is protected, and the safety of the information in the transmission process is improved.
Description of the drawings:
the technical scheme of the invention is further explained by combining the attached drawings.
Fig. 1 is a flowchart of an image data hiding method based on tensor decomposition according to an embodiment of the present application.
FIG. 2 is a flowchart illustrating the three-order cube tensor placement of an image according to an embodiment of the present application.
Fig. 3 is a flowchart of information embedding according to an embodiment of the present application.
Fig. 4 is a flowchart of information extraction according to an embodiment of the present application.
The specific implementation mode is as follows:
the objects, aspects and advantages of the present invention will be described in detail with reference to the following detailed description and accompanying drawings.
Fig. 1 is a schematic view of the entire process of the embodiment of the present application, which is specifically implemented as follows:
in step S110, the original RGB image is put into a randomly created third-order cube tensor.
And step S120, performing high-order singular value decomposition on the third-order tensor to obtain singular values and singular vectors.
Step S130, the left singular matrix element corresponding to the minimum singular value in a certain mode is replaced with the required embedded information.
And step S140, reconstructing a tensor by the replaced left singular matrix and the core tensor, and finding an image in the tensor.
And S150, obtaining the original three-order cube tensor and the image after embedding the information, and performing high-order singular value decomposition to obtain the embedded information.
The following describes some key steps of the embodiments of the present application in detail.
1. FIG. 2 shows the creation of a third order cube tensor
The embodiment of the application uses an original RGB image I O Putting the cubic tensor into a three-order cubic tensor T with random values, wherein:
original RGB image I O The size is n multiplied by m multiplied by 3;
three-order cube tensor T with random values O ;
Cubic tensor T of third order O If n > m, the tensor size is nxnxnxnxn, the corresponding mode is the first mode, if n < m, the tensor size is mxmxmxmxm, and the corresponding mode is the second mode (in this embodiment, it is assumed that n > m, the first mode is selected);
the elements of the filling are between 0 and 255; (ii) a
In third order tensor T O Randomly selecting a certain position to put in an original RGB image I O Obtaining a third order tensor T put into the image E ;
This location is a key, determined by the encryptor.
2. FIG. 3 shows a process for embedding hidden information
The embodiment of the application utilizes the minimum singular value to embed information, wherein:
step S310, using high-order singular value decomposition to the third-order tensor T E Carrying out decomposition;
performing Tucker decomposition on the tensor to obtain a core tensor and a matrix under each mode;
the Tucker decomposition process may be expressed as: t is E =[S,A,B,C]Where S is the core tensor, A, B, C are the matrices under the three modes, respectively, which is approximately equal to T E ≈S× 1 A U × 2 B U × 3 C U . Respectively carrying out SVD decomposition on A, B and C to obtain singular matrixes under each mode:
[A U ,A S ,A V ]=SVD(A);
[B U ,B S ,B V ]=SVD(B);
[C U ,C S ,C V ]=SVD(C);
the subscripts U, S and V respectively represent a left singular matrix, a singular value matrix and a right singular matrix under the mode;
step S320, selecting a singular value matrix A obtained by SVD decomposition under a mode S Selecting the minimum singular value sigma min At the left singular matrix A U To find the sigma min The corresponding column vector;
step S330, replacing the column vector with the information to be embedded to obtain a left singular matrix A 'with the embedded information' U ;
Step S340, embedding information into the core tensor S and the mode I to form a left singular matrix A' U And left singular matrix B in mode two and mode three U ,C U Obtaining a third-order tensor after information embedding;
extracting information-embedded image I from reconstructed third-order tensor by using position key R 。
3. FIG. 4 shows a process for embedded information extraction
Step S410, using the position key to embed the image I of the information R Put the initial third order cube tensor T O Obtain a new third order tensor T N ;
Step S420, for T N Performing high-order singular value decomposition (and for T) E The same decomposition is done) to obtain a core tensor S 2 And the singular matrixes under each mode comprise three modes and nine singular matrixes;
step S430, finding a left singular matrix under the mode one, wherein the singular vectors of the row corresponding to the minimum singular value are embedded information;
step S440, after extracting the embedded information, the core tensor S 2 Reconstructing a third order tensor T by using the left singular matrix extracted with information in the first mode and the corresponding left singular matrices in the other two modes K ;
Step S450, using the position key, the third order tensor T after reconstruction K Extracting the image to obtain the image with the embedded information, the image quality and the originalThe starting image quality is similar.
Claims (5)
1. An image data hiding method based on tensor decomposition, which is used for transmitting information through an image, and is characterized by comprising the following steps:
establishing a third-order cube tensor with an element value between 0 and 255, and putting an original RGB image into the third-order cube tensor;
performing Tucker decomposition on the third-order cube tensor by using high-order singular value decomposition to obtain a core tensor and singular matrixes under three modes, namely:
T E =[S,A,B,C]
wherein, T E Representing a third-order cube tensor to be decomposed, wherein S is a decomposed core tensor, and A, B and C are singular matrixes under three modes;
SVD decomposition is carried out on the singular matrixes under each mode respectively to obtain a left singular matrix, a singular value matrix and a right singular matrix under the mode:
[A U ,A S ,A V ]=SVD(A)
[B U ,B S ,B V ]=SVD(B)
[C U ,C S ,C V ]=SVD(C)
subscripts U, S and V respectively represent a left singular matrix, a singular value matrix and a right singular matrix in the mode;
selecting a singular value matrix obtained by decomposition in a first mode to find a left singular matrix element corresponding to the minimum singular value, and replacing the column of elements with the information to be embedded to obtain a new left singular matrix in the mode;
reconstructing a third-order cube tensor by using a left singular matrix and a core tensor under three modes, and extracting an image after information embedding from the third-order cube tensor;
and (4) placing the image after embedding the information into the original three-order cube tensor to obtain the embedded information through high-order singular value decomposition.
2. The method for hiding image data based on tensor decomposition as recited in claim 1, wherein the original RGB image is put into a third-order cube tensor, and comprises:
putting an RGB image of the required embedded information into a third-order cube tensor;
the position of the placement is used as a key;
the value of the third-order cube tensor element is between 0 and 255;
the size is the maximum of the image size.
3. The tensor decomposition-based image data hiding method according to claim 1, wherein performing a Tucker decomposition on the third-order cube tensor by using a higher-order singular value decomposition comprises:
decomposing the three-order cube tensor put into the image by using high-order singular value decomposition;
obtaining a core tensor and singular matrixes of the core tensor on three modes, wherein the singular matrixes comprise a singular value matrix, a left singular matrix and a right singular matrix;
selecting a left singular matrix element corresponding to a minimum singular value obtained by modal decomposition corresponding to the original image size;
and replacing the column of elements with the required embedded information to obtain a left singular matrix after the information is embedded in the mode.
4. The tensor decomposition-based image data hiding method according to claim 1, wherein reconstructing an image by using the replaced left singular matrix and the core tensor comprises:
reconstructing a third-order cube tensor by using the left singular matrix in the modality and the left singular matrices in other modalities after embedding information and the core tensor;
finding out an RGB image embedded with information from the reconstructed third-order cube tensor by using the position key;
the size of the RGB image after embedding the information is consistent with that of the original RGB image.
5. The tensor decomposition-based image data hiding method as recited in claim 1, wherein the extracting of the embedded information is obtained by performing high-order singular value decomposition on a third-order cubic tensor of the image in which the embedded information is placed, and comprises the following steps:
the image embedded with the information is placed into the initial third-order cube tensor by using the position key;
performing high-order singular value decomposition on the three-order cube tensor in which the image is placed;
the embedded information is a singular vector corresponding to the smallest singular value in the modality corresponding to the original image size.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811309996.9A CN109543425B (en) | 2018-10-26 | 2018-10-26 | Image data hiding method based on tensor decomposition |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811309996.9A CN109543425B (en) | 2018-10-26 | 2018-10-26 | Image data hiding method based on tensor decomposition |
Publications (2)
Publication Number | Publication Date |
---|---|
CN109543425A CN109543425A (en) | 2019-03-29 |
CN109543425B true CN109543425B (en) | 2023-03-10 |
Family
ID=65846114
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201811309996.9A Active CN109543425B (en) | 2018-10-26 | 2018-10-26 | Image data hiding method based on tensor decomposition |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109543425B (en) |
Families Citing this family (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112311953A (en) * | 2020-09-28 | 2021-02-02 | 北京临近空间飞行器系统工程研究所 | Reversible image information hiding method and device, electronic equipment and medium |
CN116912107B (en) * | 2023-06-13 | 2024-04-16 | 万基泰科工集团数字城市科技有限公司 | DCT-based weighted adaptive tensor data completion method |
Citations (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2011026365A1 (en) * | 2009-09-03 | 2011-03-10 | 中兴通讯股份有限公司 | Method and system for embedding and extracting image digital watermark |
WO2014071714A1 (en) * | 2012-11-06 | 2014-05-15 | 中国科学院自动化研究所 | Watermark embedding and extracting method |
AU2017100244A4 (en) * | 2017-02-13 | 2017-03-30 | Macau University Of Science And Technology | Methods and Apparatus for Color Image Watermarking |
CN106791872A (en) * | 2016-11-18 | 2017-05-31 | 南京邮电大学 | Information concealing method based on SVD |
CN107247688A (en) * | 2017-06-02 | 2017-10-13 | 上海德衡数据科技有限公司 | A kind of multichannel data compressed sensing cube framework |
CN107274333A (en) * | 2017-06-21 | 2017-10-20 | 鲁东大学 | A kind of spatial domain coloured image blind watermark method for merging singular value decomposition |
CN107292806A (en) * | 2017-06-28 | 2017-10-24 | 南京师范大学 | A kind of remote sensing image digital watermark embedding and extracting method based on quaternion wavelet |
CN107835427A (en) * | 2017-11-29 | 2018-03-23 | 福建北卡科技有限公司 | A kind of color image information hidden method based on contourlet and SVD |
CN108681982A (en) * | 2018-04-04 | 2018-10-19 | 微梦创科网络科技(中国)有限公司 | A kind of embedding algorithm and extracting method and system based on singular value decomposition |
Family Cites Families (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US7787654B2 (en) * | 2007-09-21 | 2010-08-31 | Shibaura Institute Of Technology | Electronic watermark embedding apparatus and method and electronic watermark detecting apparatus and method |
-
2018
- 2018-10-26 CN CN201811309996.9A patent/CN109543425B/en active Active
Patent Citations (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2011026365A1 (en) * | 2009-09-03 | 2011-03-10 | 中兴通讯股份有限公司 | Method and system for embedding and extracting image digital watermark |
WO2014071714A1 (en) * | 2012-11-06 | 2014-05-15 | 中国科学院自动化研究所 | Watermark embedding and extracting method |
CN106791872A (en) * | 2016-11-18 | 2017-05-31 | 南京邮电大学 | Information concealing method based on SVD |
AU2017100244A4 (en) * | 2017-02-13 | 2017-03-30 | Macau University Of Science And Technology | Methods and Apparatus for Color Image Watermarking |
CN107247688A (en) * | 2017-06-02 | 2017-10-13 | 上海德衡数据科技有限公司 | A kind of multichannel data compressed sensing cube framework |
CN107274333A (en) * | 2017-06-21 | 2017-10-20 | 鲁东大学 | A kind of spatial domain coloured image blind watermark method for merging singular value decomposition |
CN107292806A (en) * | 2017-06-28 | 2017-10-24 | 南京师范大学 | A kind of remote sensing image digital watermark embedding and extracting method based on quaternion wavelet |
CN107835427A (en) * | 2017-11-29 | 2018-03-23 | 福建北卡科技有限公司 | A kind of color image information hidden method based on contourlet and SVD |
CN108681982A (en) * | 2018-04-04 | 2018-10-19 | 微梦创科网络科技(中国)有限公司 | A kind of embedding algorithm and extracting method and system based on singular value decomposition |
Non-Patent Citations (3)
Title |
---|
Unitary Embedding for Data Hiding with the SVD;Clifford Bergman et al.;《Security,Strganography,and Watermarking of Multimedia Contents VII》;20050321;全文 * |
基于分块DCT和Tucker分解的彩色图像数字水印;严明等;《计算机应用》;20100601(第06期);全文 * |
基于小波变换和Tucker分解的彩色图像数字水印;夏建平等;《计算机应用研究》;20100215(第02期);全文 * |
Also Published As
Publication number | Publication date |
---|---|
CN109543425A (en) | 2019-03-29 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Anand et al. | Watermarking techniques for medical data authentication: a survey | |
Lin et al. | Invertible secret image sharing with steganography | |
Qin et al. | An efficient coding scheme for reversible data hiding in encrypted image with redundancy transfer | |
Naseri et al. | A new secure quantum watermarking scheme | |
Lu et al. | Dual imaging-based reversible hiding technique using LSB matching | |
Prasad et al. | An RGB colour image steganography scheme using overlapping block-based pixel-value differencing | |
Chan et al. | Hiding data in images by simple LSB substitution | |
Yu et al. | A new steganographic method for color and grayscale image hiding | |
CN106791872B (en) | Information concealing method based on SVD | |
GNDU RC | Dual layer security of data using LSB image steganography method and AES encryption algorithm | |
Chang et al. | A sudoku-based secret image sharing scheme with reversibility | |
CN113115053B (en) | Image encryption method based on integer wavelet transform and compressed sensing | |
Xiang et al. | Cryptanalysis and improvement of a reversible data-hiding scheme in encrypted images by redundant space transfer | |
CN109543425B (en) | Image data hiding method based on tensor decomposition | |
CN106611399A (en) | Information hiding algorithm specific to encrypted image in cloud environment | |
CN109064381B (en) | Multi-dimensional digital watermark encryption method based on secret key | |
Lee et al. | A new steganographic method based on information sharing via PNG images | |
Han et al. | DWT-domain dual watermarking algorithm of color image based on visual cryptography | |
CN110706145B (en) | Reversible remote sensing image information hiding processing method and device | |
Sengupta et al. | Self authentication of color images through discrete cosine transformation (SADCT) | |
Wang et al. | A novel (n, n) secret image sharing scheme based on Sudoku | |
Tulasidasu et al. | Steganography based secret image sharing using block division technique | |
Anumol et al. | Enhanced security in medical image steganography-a hybrid approach using spatial and transform domain | |
CN109920015B (en) | Image visual meaningful encryption method based on fractal graph generation | |
Ye et al. | An image hiding scheme based on 3D skew tent map and discrete wavelet transform |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |