CN113706364A - Reversible information hiding method for remote sensing image - Google Patents
Reversible information hiding method for remote sensing image Download PDFInfo
- Publication number
- CN113706364A CN113706364A CN202111074268.6A CN202111074268A CN113706364A CN 113706364 A CN113706364 A CN 113706364A CN 202111074268 A CN202111074268 A CN 202111074268A CN 113706364 A CN113706364 A CN 113706364A
- Authority
- CN
- China
- Prior art keywords
- carrier
- weight
- picture set
- remote sensing
- image
- 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.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 32
- 230000002441 reversible effect Effects 0.000 title claims abstract description 26
- 238000013528 artificial neural network Methods 0.000 claims abstract description 30
- 238000012549 training Methods 0.000 claims abstract description 27
- 230000001502 supplementing effect Effects 0.000 claims abstract description 5
- 210000002569 neuron Anatomy 0.000 claims description 13
- 238000012545 processing Methods 0.000 claims description 5
- 230000000295 complement effect Effects 0.000 claims description 4
- 230000008569 process Effects 0.000 claims description 4
- 238000004364 calculation method Methods 0.000 claims description 3
- 230000008859 change Effects 0.000 claims description 3
- 238000010276 construction Methods 0.000 claims description 2
- 238000010586 diagram Methods 0.000 description 5
- 238000013475 authorization Methods 0.000 description 3
- 238000004891 communication Methods 0.000 description 3
- 238000007781 pre-processing Methods 0.000 description 3
- 238000011160 research Methods 0.000 description 3
- 230000009466 transformation Effects 0.000 description 3
- 230000003044 adaptive effect Effects 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 230000006872 improvement Effects 0.000 description 2
- 241000233855 Orchidaceae Species 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 238000000354 decomposition reaction Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000009189 diving Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000005284 excitation Effects 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 230000004927 fusion Effects 0.000 description 1
- 230000007774 longterm Effects 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 238000003672 processing method Methods 0.000 description 1
- 238000013139 quantization Methods 0.000 description 1
- 230000005855 radiation Effects 0.000 description 1
- 230000008707 rearrangement Effects 0.000 description 1
- 238000011084 recovery Methods 0.000 description 1
- 238000012163 sequencing technique Methods 0.000 description 1
- 238000010008 shearing Methods 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T1/00—General purpose image data processing
- G06T1/0021—Image watermarking
- G06T1/005—Robust watermarking, e.g. average attack or collusion attack resistant
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/22—Matching criteria, e.g. proximity measures
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/048—Activation functions
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N1/00—Scanning, transmission or reproduction of documents or the like, e.g. facsimile transmission; Details thereof
- H04N1/32—Circuits or arrangements for control or supervision between transmitter and receiver or between image input and image output device, e.g. between a still-image camera and its memory or between a still-image camera and a printer device
- H04N1/32101—Display, printing, storage or transmission of additional information, e.g. ID code, date and time or title
- H04N1/32144—Display, printing, storage or transmission of additional information, e.g. ID code, date and time or title embedded in the image data, i.e. enclosed or integrated in the image, e.g. watermark, super-imposed logo or stamp
- H04N1/32149—Methods relating to embedding, encoding, decoding, detection or retrieval operations
- H04N1/32267—Methods relating to embedding, encoding, decoding, detection or retrieval operations combined with processing of the image
- H04N1/32272—Encryption or ciphering
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N1/00—Scanning, transmission or reproduction of documents or the like, e.g. facsimile transmission; Details thereof
- H04N1/32—Circuits or arrangements for control or supervision between transmitter and receiver or between image input and image output device, e.g. between a still-image camera and its memory or between a still-image camera and a printer device
- H04N1/32101—Display, printing, storage or transmission of additional information, e.g. ID code, date and time or title
- H04N1/32144—Display, printing, storage or transmission of additional information, e.g. ID code, date and time or title embedded in the image data, i.e. enclosed or integrated in the image, e.g. watermark, super-imposed logo or stamp
- H04N1/32149—Methods relating to embedding, encoding, decoding, detection or retrieval operations
- H04N1/32347—Reversible embedding, i.e. lossless, invertible, erasable, removable or distorsion-free embedding
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Computing Systems (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Multimedia (AREA)
- Signal Processing (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Editing Of Facsimile Originals (AREA)
- Image Processing (AREA)
Abstract
The invention discloses a reversible information hiding method of remote sensing images, which comprises the steps of obtaining a first weight carrier image and a non-weight carrier image from an original remote sensing image, and respectively splitting the first weight carrier image and the non-weight carrier image into a first original remote sensing image set, a weight carrier image set and a non-weight carrier image set; splitting or supplementing a carrier picture set, obtaining a second original remote sensing picture set which is mapped with a final carrier picture set one by adopting an SIFT value judgment method, and training an initialized perceptron network by adopting a back propagation algorithm to obtain a perceptron neural network and a weight file; and selecting a weight value area and a plurality of weight information from the original remote sensing image by utilizing a plurality of encryption algorithms, and embedding the encrypted weight information into the weight value area by utilizing a robust digital watermarking algorithm to finally obtain the remote sensing image with the hidden reversible information. The method can ensure the safety and has higher reversibility.
Description
Technical Field
The invention belongs to the technical field of remote sensing, and particularly relates to a reversible information hiding method for remote sensing images.
Background
The existing image information hiding algorithms are various, but most of the existing image information hiding algorithms only aim at common images, and compared with the common images, the remote sensing images have three different points that the first image is large in size; secondly, the texture structure is complex; the requirement of the user on the data precision is high, and when the hidden information is extracted, the original data needs to be accurately restored. The information hiding algorithm of the common image is difficult to process images with large size and rich structure, and a receiver loses great precision when extracting an original image, so that the use of the image is influenced, and the information hiding algorithm aiming at the remote sensing image must be specially researched.
With the improvement of the remote sensing image information hiding algorithm technology, expert scholars begin to pay attention to the application of the information hiding algorithm in the remote sensing image. Foreign scholars have been concerned about the research of digital watermarking, so that the research of digital watermarking technology in the field is quite mature, but an information hiding algorithm for spatial confidential information is still in the beginning. In China, a plurality of articles hidden for remote sensing image information are also appeared recently. The definition of information hiding is expanded by Wangxian and Guanze groups: the original information hiding is to embed the confidential information into a carrier irrelevant to the confidential information, and the new algorithm is to take an artifact image after the confidential information is extracted as the carrier, so that the transmission of data volume is saved, and the security of the confidential information is improved. Meanwhile, an airspace algorithm for hiding the remote sensing information is provided, the algorithm realizes the information hiding through a JPEG quantization table and an odd-even diving method, and the algorithm has no influence on an original remote sensing image. Wangxian and the like put forward the authorization use of confidential information for the first time, and simultaneously apply wavelet transformation, color space transformation and Singular Value Decomposition (SVD) to image fusion respectively, thereby obtaining a new pseudo remote sensing image generation algorithm and realizing the authorization use of the confidential information. The Wanglan provides a reversible DCT domain information hiding algorithm which is suitable for remote sensing images with high radiation resolution, the algorithm realizes the embedding of confidential information by modifying high-frequency coefficients after DCT transformation, the algorithm can better resist linear stretching attack, and meanwhile, the algorithm has good stability for common attacks such as shearing, noise, filtering and the like. The royal orchid and the like also provide a lossless LSB hiding algorithm for the remote sensing image, the algorithm has larger confidential information embedding amount, can realize lossless recovery of information, and can be used for authorization use of the confidential information.
The information hiding algorithm specially aiming at the remote sensing image comprises a discrete wavelet domain digital watermarking algorithm, a discrete cosine transform algorithm, a spatial information hiding matrix method, a derived human visual system embedding strategy (improvement of discrete wavelets) based on a discrete wavelet domain, a pixel replacement algorithm, a latest high-capacity embedding algorithm, an adaptive hiding algorithm and a remote sensing image multi-coverage hiding and writing algorithm which are proposed in 2017.
The algorithms have substantially the same steps of determining an encrypted area in an original image, selecting a proper carrier image according to the property of the encrypted area image, and finally embedding the encrypted area image into the carrier image through respective algorithms. The key point of the research of the algorithms is that the method of selecting the carrier is adopted; ② the way of embedding information.
The disadvantage of this kind of algorithm is that it is difficult to find a suitable carrier image, and even with an adaptive algorithm (an algorithm in which a user can select a carrier autonomously), the security and imperceptibility of the algorithm are related to the user's selection. Secondly, reversibility and imperceptibility are difficult to balance when information is embedded, the algorithm idea is usually focused on information hiding and neglects the process of information extraction of a receiver, and therefore the reversibility degree is low. In order to solve the problem of reversible information hiding, many scholars propose a reversible information hiding algorithm, which improves the precision of images extracted by receivers, but reduces the safety of information.
Disclosure of Invention
The invention discloses a reversible information hiding method for remote sensing images, which can ensure the safety and has higher reversibility.
A reversible information hiding method for remote sensing images comprises the following steps:
s1: selecting a first carrier image from the original remote sensing image, dividing the first carrier image into a first weight carrier image and a non-weight carrier image, splitting the original remote sensing image to obtain a first original remote sensing picture set, splitting the first weight carrier image to obtain a weight carrier picture set, and splitting the non-weight carrier image to obtain a non-weight carrier picture set;
s2: combining the weighted carrier picture set and the non-weighted carrier picture set to obtain a carrier picture set, splitting or supplementing the carrier picture set to obtain a final carrier picture set with the same number of pictures as the original remote sensing picture set, indexing pictures with the lowest similarity in the first original remote sensing picture set for each picture in the final carrier picture set by adopting an SIFT value judgment method to obtain a second original remote sensing picture set which is mapped with the final carrier picture set one by one, and taking the final carrier picture set as a training sample set;
s3: training an initialized perceptron network by adopting a back propagation algorithm based on a training sample set to obtain a perceptron neural network, and constructing a weight file of weights of a plurality of neurons in the perceptron neural network;
s4: selecting a second carrier image from the original remote sensing image, selecting a second weight carrier image from the second carrier image by using a first encryption algorithm, selecting a first weight area from the second weight carrier image by using a second encryption algorithm, dividing a weight file into a plurality of weight information, encrypting the plurality of weight information by using a third encryption algorithm, embedding the encrypted weight information into the first weight area by using a robust digital watermarking algorithm, and sending the first, second and third encryption algorithms and the robust digital watermarking algorithm to a user to finally obtain the remote sensing image with hidden reversible information.
The invention carries out preprocessing on the carrier image, namely splitting, complementing color and recombining (because weight information is very small relative to the carrier file, and each carrier file is not necessary to carry the weight information, the preprocessing is carried out), a final carrier picture set is used as a training sample set and is input to an original remote sensing picture set output by an initial perceptron neural network to be compared with a second original remote sensing picture set, the training is completed when a threshold value is met, and a picture set with the lowest similarity to the final carrier picture set is used as an output standard of the perceptron neural network, so that the problem that similar pictures cannot be distinguished is avoided, and the purpose of accurately obtaining the original image is achieved.
According to the invention, the weight area and the weight file are hidden through triple encryption, so that a protection effect is achieved, and even if the weight area and the weight information are found, the weight information is protected by the robust digital watermarking algorithm, so that the weight information is prevented from being maliciously modified.
According to the invention, the first, second and third encryption algorithms and the robust digital watermarking algorithm are sent to the user, so that the user can restore the file to obtain the original image based on the algorithm in a reversible way, and the encryption and watermarking algorithms can be decided by the user, so that the method has better flexibility.
The user decrypts the remote sensing image with the hidden reversible information through a robust digital watermarking algorithm to obtain a plurality of encrypted weight information, decrypts the remote sensing image through a third encryption algorithm to obtain a plurality of weight information, combines the weight information to obtain a weight file, decrypts the first weight area through a second encryption algorithm to obtain a second weight carrier image, decrypts the second weight carrier image through the first encryption algorithm to obtain a second carrier image, obtains a perceptron neural network through the weight file, and inputs the second carrier image to the perceptron neural network to obtain an original remote sensing image.
Splitting the first weight carrier image to obtain a weight carrier image set, comprising:
hollowing out a second weight area in the weight carrier image, performing color complementation on the hollowed second weight area, splitting the color complemented weight carrier image to obtain an initial weight carrier picture set, and performing color complementation on a redundant area in the initial weight carrier picture set again to obtain the weight carrier picture set.
Splitting the non-weighted carrier image to obtain a non-weighted carrier image set, comprising:
and splitting the unweighted carrier image to obtain an initial unweighted carrier picture set, and performing complementary color on redundant areas in the initial unweighted carrier picture set to obtain an unweighted carrier picture set.
And numbering the pictures in the weighted carrier picture set and the pictures in the non-weighted carrier picture set respectively, and sequencing the pictures in the combined carrier picture set according to the numbering result.
Splitting or supplementing the carrier picture set to obtain a final carrier picture set with the same number of pictures as the original remote sensing picture set, wherein the final carrier picture set comprises the following steps:
if the number of pictures in the carrier picture set is less than that of the original remote sensing picture set, the RGB value is (beta)k+i mod 2 56,βk+i mod 2 56，βk+ i mod 256) image is supplemented to the carrier picture set, so that the number of pictures in the supplemented carrier picture set is consistent with the number of pictures in the original remote sensing picture set, where k is 1,2, … … β, and β is the number of carrier pictures, and it can be ensured that each carrier picture can correspond to one original picture during training, so that the carrier image and the original image are split into the same number of sub-images.
And if the number of the pictures in the carrier picture set is greater than that of the original remote sensing picture set, deleting the number of the pictures in the carrier picture set, so that the number of the pictures in the supplemented carrier picture set is consistent with that of the pictures in the original remote sensing picture set.
The initial perceptron neural network comprises an input layer, a plurality of hidden layers and an output layer, wherein the input layer is used for receiving a final carrier picture set and inputting the final carrier picture set into the hidden layers, the hidden layers are used for processing the final carrier picture by adopting a back propagation algorithm to obtain a third original remote sensing picture set, when the error between the third original remote sensing picture set obtained by utilizing a loss function and the second original remote sensing picture set meets a threshold value, the back propagation calculation is stopped to obtain the final original remote sensing picture set, and meanwhile training of the initial perceptron network is completed to obtain the perceptron neural network.
And in the process of training the initial perceptron neural network by using a back propagation algorithm, optimizing the trained initial perceptron neural network by using a BN algorithm to obtain the perceptron neural network.
The weight construction method of the weight of a plurality of neurons in the perceptron neural network comprises the following steps:
the initial perceptron neural network comprises a plurality of neurons, an initial value is set for the weight of each neuron, and after training is completed, the weight change value set of each neuron is used as a weight file.
Compared with the prior art, the invention has the beneficial effects that:
the reversible information hiding method for the remote sensing image has higher safety, does not depend on the image independently selected by a user, does not need the user to determine a specific encryption area, and has high safety when being embedded as a digital watermark, and a weight file for decryption of a receiver is very small relative to a carrier image.
Although the time required in the step of hiding the image is longer than that of other algorithms, the requirement on the precision is far longer than that on the time when the problems of satellite images and the like are processed, so that the defect in the time is acceptable.
The invention has strong flexibility, and the receiver and the sender can freely adjust the training times and the number of samples of the perceptron according to the requirements, freely select the encryption algorithm and the digital watermarking algorithm when in embedding, and apply to different ranges according to different requirements of users.
Drawings
Fig. 1 is a diagram illustrating a method for converting a single-band remote sensing image into an RGB format according to an embodiment;
FIG. 2 is a diagram illustrating a method for converting a multiband remote sensing image into an RGB format according to an embodiment;
FIG. 3 is a diagram illustrating an RGB format image splitting of an unweighted carrier according to an embodiment;
FIG. 4 is a diagram illustrating a weight carrier RGB format image splitting diagram according to an embodiment;
FIG. 5 is a flowchart of image sorting according to an embodiment;
fig. 6 is a flowchart of embedding weight information according to an embodiment.
Detailed Description
The technical scheme adopted by the invention for solving the technical problem specifically comprises the following steps:
s1: the image preprocessing comprises the following specific steps:
as shown in fig. 1, the single-band remote sensing image is a gray scale image, the gray scale image is directly converted into an RGB image (i.e. the values of the other two channels are compensated to 0), as shown in fig. 2, the multi-band remote sensing image is sequentially split into a plurality of single-band gray scale images according to the bands, a false color RGB remote sensing image (TIFF) is synthesized for every three single bands, and the false color RGB remote sensing image (TIFF) is used as the original remote sensing image.
N different first carrier images { p _1, p _2 … … p _ n } are selected from the original remote sensing image, and must be meaningful images, and the image size needs to be the same as or similar to the original image. The image processing method can be used for processing the remote sensing image without the secret value into an RGB format or an ordinary RGB picture, and m images are selected from the n carrier images to be used as first weight carrier images
Wherein i is an index, is specially used for embedding weight information (not all carrier images need to be embedded with weight information), and selects a part of the n carrier images as non-weight carrier images
As shown in fig. 3, the unweighted carrier image is divided into a plurality of pictures with 200 × 200 pixels, and the serial number information is recorded as
Performing complementary color processing on redundant areas of a plurality of split pictures with pixels of 200 multiplied by 200, wherein RBG values of the partial areas are set as (i mod 256, i mod 256, i mod 256);
as shown in fig. 4, splitting the first weight carrier image to obtain a weight carrier picture set: selecting weight area, hollowing out, performing color compensation, and drawing
The RGB value of the hollow area is set to (i mod 256, i mod 256, i mod 256). Then splitting the picture, and recording sequence number information:
and (3) performing complementary color processing on redundant areas (namely gray areas in FIG. 4) of the split pictures, wherein the RBG values of all the areas are set to be (i mod 256, i mod 256, i mod 256).
Ordering n original images and recording sequence number information { q1，q2…qn}. Splitting according to the mode of fig. 3, and recording sequence number information: { q ] q11，q12，…q21，q22，…qn1，qn2…}。
S2: the image reordering comprises the following specific steps:
combining the weighted carrier picture set and the non-weighted carrier picture set to obtain a carrier picture set pjiSorting j and i in order from small to large respectively: { p11，p12，p13…p21，p22…pn1，pn2…}；
As shown in fig. 5, the carrier picture set is split or supplemented to obtain a final carrier picture set with the same number of pictures as the original remote sensing picture set, and if the number of pictures in the carrier picture set is less than that of the original remote sensing picture set, the RGB value is (β)k+i mod 2 56，βk+i mod 2 56，βk+ i mod 256) image is supplemented to the carrier picture set, so that the number of pictures in the supplemented carrier picture set is consistent with the number of pictures in the original remote sensing picture set, where k is 1,2, … … β, and β is the number of carrier pictures, and it can be ensured that each carrier picture can correspond to one original picture during training, so that the carrier image and the original image are split into the same number of sub-images.
And if the number of the pictures in the carrier picture set is greater than that of the original remote sensing picture set, deleting the number of the pictures in the carrier picture set, so that the number of the pictures in the supplemented carrier picture set is consistent with that of the pictures in the original remote sensing picture set.
Indexing a picture with the lowest similarity in a first original remote sensing picture set for each picture in a final carrier picture set by adopting an SIFT value judgment method to obtain a second original remote sensing picture set which is mapped with the final carrier picture set one by one, and taking the final carrier picture set as a training sample set, wherein the method specifically comprises the following steps: rearrangement and serial number update are carried out to order the carrier image p11Selecting a graph with the lowest similarity (using Scale artifact Feature Transform) in the original image as a starting point, wherein the similarity is judged by using a Scale artifact Feature Transform (note: SIFT (A, B) ═ similarity between B graph and A graph, the closer to 1 the value indicates that B is more similar to A, and the closer to 0 the value indicates that B is more dissimilar to A), and selecting a graph with the lowest similarity from the original image as the starting point
Then using the carrier image p12Repeating the above steps for the starting point to find
By analogy, rearranging the original image, and recording the sequence number information:
finally, two rows of carrier images and original images with the same quantity are generated, so that
Mapping one by one to make samples
For the training of the perceptron.
S3: training and optimizing a network initialization sensor network, and generating a weight file:
training an initialized perceptron network by adopting a back propagation algorithm based on a training sample set to obtain a perceptron neural network, and constructing a weight file of weights of a plurality of neurons in the perceptron neural network; let m mean the number of hidden layer nodes and n mean the number of inputThe number of nodes in the input layer, l, the number of nodes in the output layer, and a constant between (0 and 1). According to empirical formulas:
since the input and output are both 200 × 200 pictures, i.e. 40000 neurons, i.e. n is 40000, m is 285, i.e. the optimal number of nodes of the hidden layer is 285, plus four offset variables. A multilayer perceptron neural network is arranged, and the multilayer perceptron neural network comprises an input layer, three hidden layers and an output layer, wherein 7698528 weights are provided. Setting the initial weight values to be 0, wherein excitation functions of all layers are ReLU (Rectified Linear Unit);
continuously training a network by using a BP (back propagation) algorithm for the initialized perceptron, stopping back propagation calculation to obtain a final original remote sensing picture set when the error between a third original remote sensing picture set obtained by using a loss function and a second original remote sensing picture set meets a threshold value, finishing training the initialized perceptron network to obtain a perceptron neural network, and taking a weight value change value set of each neuron as a weight value file after training is finished;
s4: as shown in fig. 6, the weight file is divided and embedded, and the specific steps are as follows:
selecting a second carrier image from an original remote sensing image, selecting a second weight carrier image from the second carrier image by using a first encryption algorithm, selecting a first weight area from the second weight carrier image by using a second encryption algorithm, dividing a weight file into a plurality of weight information, encrypting the plurality of weight information by using a third encryption algorithm, embedding the encrypted weight information into the first weight area by using a robust digital watermarking algorithm, and sending the first, second and third encryption algorithms and the robust digital watermarking algorithm to a user to finally obtain the remote sensing image with hidden reversible information, wherein the specific steps are as follows:
in n second carrier images, using an encryption algorithm 1 (algorithm selected by user), extracting x images as second weight carrier images, and transmitting a secret key 1 to a receiver before communication for long-term use.
In the selected x second weight carrier images, an encryption algorithm 2 (selected by a user) is respectively used, a partial area is extracted as a first weight area, and a secret key 2 is transmitted to a receiver before communication so as to be used for a long time.
Let the finally generated weight file be M bit, and equally divide it into d parts (d is min { x, M }, and x represents the extracted x second weight carrier images) of multiple weight information. Respectively embedded in the second weight carrier images. Before storing the weight information, the weight information after the peer division is encrypted by using an encryption algorithm 3 (the algorithm is selected by a user), and the secret key 3 is transmitted to a receiver before communication so as to be used for a long time.
And embedding the encrypted weight information into a designated weight area in a weight carrier by using a robust digital watermarking algorithm (an algorithm is selected by a user), wherein the weight carrier image does not need to be preprocessed but is directly transmitted after information is embedded, and the weight carrier image needing to be preprocessed is only a sample input into a perceptron for training. The encryption algorithms 1,2, 3 and the robust digital watermarking algorithm can be decided by the sender. The purpose of selecting the robust digital watermarking algorithm for embedding is that weight information cannot be lost even if an attacker knows the position of the weight information and maliciously modifies the weight information (as long as the weight carrier image is not modified to the extent that the weight carrier image cannot be used at all).
Claims (9)
1. A reversible information hiding method for remote sensing images is characterized by comprising the following steps:
s1: selecting a first carrier image from the original remote sensing image, dividing the first carrier image into a first weight carrier image and a non-weight carrier image, splitting the original remote sensing image to obtain a first original remote sensing picture set, splitting the first weight carrier image to obtain a weight carrier picture set, and splitting the non-weight carrier image to obtain a non-weight carrier picture set;
s2: combining the weighted carrier picture set and the non-weighted carrier picture set to obtain a carrier picture set, splitting or supplementing the carrier picture set to obtain a final carrier picture set with the same number of pictures as the first original remote sensing picture set, indexing pictures with the lowest similarity in the first original remote sensing picture set for each picture in the final carrier picture set by adopting an SIFT value judgment method to obtain a second original remote sensing picture set which is mapped with the final carrier picture set one by one, and taking the final carrier picture set as a training sample set;
s3: training an initialized perceptron network by adopting a back propagation algorithm based on a training sample set to obtain a perceptron neural network, and constructing a weight file of weights of a plurality of neurons in the perceptron neural network;
s4: selecting a second carrier image from the original remote sensing image, selecting a second weight carrier image from the second carrier image by using a first encryption algorithm, selecting a first weight area from the second weight carrier image by using a second encryption algorithm, dividing a weight file into a plurality of weight information, encrypting the plurality of weight information by using a third encryption algorithm, embedding the encrypted weight information into the first weight area by using a robust digital watermarking algorithm, and sending the first, second and third encryption algorithms and the robust digital watermarking algorithm to a user to finally obtain the remote sensing image with hidden reversible information.
2. The reversible information hiding method of the remote sensing image according to claim 1, characterized in that a user decrypts the reversible information hidden remote sensing image through a robust digital watermarking algorithm to obtain a plurality of encrypted weight information, decrypts through a third encryption algorithm to obtain a plurality of weight information, merges the plurality of weight information to obtain a weight file, decrypts a first weight area through a second encryption algorithm to obtain a second weight carrier image, decrypts the second weight carrier image through a first encryption algorithm to obtain a second carrier image, obtains a perceptron neural network through the weight file, and inputs the second carrier image to the perceptron neural network to obtain an original remote sensing image.
3. The reversible information hiding method for remote sensing images according to claim 1, wherein splitting the first weight carrier image to obtain a weight carrier picture set comprises:
hollowing out a second weight area in the weight carrier image, performing color complementation on the hollowed second weight area, splitting the color complemented weight carrier image to obtain an initial weight carrier picture set, and performing color complementation on a redundant area in the initial weight carrier picture set again to obtain the weight carrier picture set.
4. The reversible information hiding method for remote sensing images according to claim 1, wherein splitting the unweighted carrier image to obtain an unweighted carrier picture set comprises:
and splitting the unweighted carrier image to obtain an initial unweighted carrier picture set, and performing complementary color on redundant areas in the initial unweighted carrier picture set to obtain an unweighted carrier picture set.
5. The reversible information hiding method for remote sensing images according to claim 1, characterized in that the pictures in the weighted carrier picture set and the pictures in the non-weighted carrier picture set are numbered respectively, and the pictures in the combined carrier picture set are sorted according to the numbering result.
6. The reversible information hiding method for remote sensing images according to claim 1, wherein splitting or supplementing the carrier picture set to obtain a final carrier picture set with the same number of pictures as the original remote sensing picture set comprises:
if the number of pictures in the carrier picture set is less than that of the original remote sensing picture set, the RGB value is (beta)k+imod256,βk+imod 256，βk+ imod 256) image is supplemented to the carrier picture set, so that the number of pictures in the supplemented carrier picture set is consistent with the number of pictures in the original remote sensing picture set, wherein k is 1,2, … … beta, and beta is the number of carrier pictures needing to be supplemented;
and if the number of the pictures in the carrier picture set is greater than that of the original remote sensing picture set, deleting the number of the pictures in the carrier picture set, so that the number of the pictures in the supplemented carrier picture set is consistent with that of the pictures in the original remote sensing picture set.
7. The reversible information hiding method for the remote sensing images according to claim 1, characterized in that the initial perceptron neural network comprises an input layer, a plurality of hidden layers and an output layer, the input layer is used for receiving the final carrier picture set and inputting the final carrier picture set to the hidden layers, the hidden layers are used for processing the final carrier picture by using a back propagation algorithm to obtain a third original remote sensing picture set, when an error between the third original remote sensing picture set obtained by using the loss function and the second original remote sensing picture set meets a threshold value, the back propagation calculation is stopped to obtain the final original remote sensing picture set, and meanwhile, training of the initialized perceptron network is completed to obtain the perceptron neural network.
8. The reversible information hiding method for remote sensing images of claim 7, wherein in the process of training the initial perceptron neural network by the back propagation algorithm, the trained initial perceptron neural network is optimized by the BN algorithm to obtain the perceptron neural network.
9. The reversible information hiding method for remote sensing images according to claim 1, wherein the construction of the weight file by the weights of a plurality of neurons in the perceptron neural network comprises:
the initial perceptron neural network comprises a plurality of neurons, an initial value is set for the weight of each neuron, and after training is completed, the weight change value set of each neuron is used as a weight file.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111074268.6A CN113706364A (en) | 2021-09-14 | 2021-09-14 | Reversible information hiding method for remote sensing image |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111074268.6A CN113706364A (en) | 2021-09-14 | 2021-09-14 | Reversible information hiding method for remote sensing image |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN113706364A true CN113706364A (en) | 2021-11-26 |
Family
ID=78660350
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111074268.6A Pending CN113706364A (en) | 2021-09-14 | 2021-09-14 | Reversible information hiding method for remote sensing image |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113706364A (en) |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101640779A (en) * | 2008-07-31 | 2010-02-03 | 鸿富锦精密工业（深圳）有限公司 | Encryption system and encryption method of image intake device |
| CN101882296A (en) * | 2010-05-07 | 2010-11-10 | 邹阿金 | Method for hiding Chebyshev neural network chaotic digital watermark information |
| CN102607533A (en) * | 2011-12-28 | 2012-07-25 | 中国人民解放军信息工程大学 | Block adjustment locating method of linear array CCD (Charge Coupled Device) optical and SAR (Specific Absorption Rate) image integrated local area network |
| WO2018214195A1 (en) * | 2017-05-25 | 2018-11-29 | 中国矿业大学 | Remote sensing imaging bridge detection method based on convolutional neural network |
-
2021
- 2021-09-14 CN CN202111074268.6A patent/CN113706364A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101640779A (en) * | 2008-07-31 | 2010-02-03 | 鸿富锦精密工业（深圳）有限公司 | Encryption system and encryption method of image intake device |
| CN101882296A (en) * | 2010-05-07 | 2010-11-10 | 邹阿金 | Method for hiding Chebyshev neural network chaotic digital watermark information |
| CN102607533A (en) * | 2011-12-28 | 2012-07-25 | 中国人民解放军信息工程大学 | Block adjustment locating method of linear array CCD (Charge Coupled Device) optical and SAR (Specific Absorption Rate) image integrated local area network |
| WO2018214195A1 (en) * | 2017-05-25 | 2018-11-29 | 中国矿业大学 | Remote sensing imaging bridge detection method based on convolutional neural network |
Non-Patent Citations (1)
| Title |
|---|
| 施垚等: "《基于多层感知机的遥感影像可逆信息隐藏算法》", 《杭州师范大学学报（自然科学版）》, vol. 18, no. 1, pages 1 - 8 * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Kumar et al. | Secure video communication using firefly optimization and visual cryptography | |
| CN108537716B (en) | Color image encryption embedding method based on discrete domain | |
| Yu et al. | A new steganographic method for color and grayscale image hiding | |
| Manjula et al. | A novel hash based least significant bit (2-3-3) image steganography in spatial domain | |
| Jenifer et al. | LSB approach for video steganography to embed images | |
| Verma et al. | Hybrid mode of medical image watermarking to enhance robustness and imperceptibility | |
| Houssein et al. | An image steganography algorithm using haar discrete wavelet transform with advanced encryption system | |
| Moosazadeh et al. | A new robust color digital image watermarking algorithm in DCT domain using genetic algorithm and coefficients exchange approach | |
| GNDU RC | State-of-the-art Review on Steganographic Techniques | |
| CN111161124B (en) | Image steganography method of reference matrix model based on mean square error evaluation | |
| Plata et al. | Robust spatial-spread deep neural image watermarking | |
| Wang et al. | TPE-ISE: approximate thumbnail preserving encryption based on multilevel DWT information self-embedding | |
| Ansari et al. | JPEG Image Steganography based on Coefficients Selection and Partition | |
| Fu et al. | Efficient capacity-distortion reversible data hiding based on combining multipeak embedding with local complexity | |
| CN109194846B (en) | EMD (n, m, delta) self-adaptive image steganography method based on complexity | |
| CN113706364A (en) | Reversible information hiding method for remote sensing image | |
| CN109543425B (en) | Image data hiding method based on tensor decomposition | |
| Yasasvy et al. | A hybrid blind watermarking with redundant discrete wavelet and Hadamard transform | |
| Manoharan et al. | Pixel value differencing method based on CMYK colour model | |
| Lakshmi Sirisha et al. | Steganography based image sharing with reversibility | |
| Klim | Selected least significant bit approach for hiding information inside color image steganography by using magic square | |
| Leelavathy et al. | Oblivious Image Watermarking in Discrete Multiwavelet Domain using QIMM. | |
| Reddy et al. | Wavelet based secure steganography with scrambled payload | |
| Revathi | Security And Preservation For Video Data Transmission Using Blockchain Technology For Cloud Storage On Internet Of Things (Iot) | |
| Naveen et al. | Image Steganography Method for Securing Multiple Images using LSB–GA |
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 |