CN114742687A - Medical image zero watermark generation algorithm based on multi-algorithm fusion - Google Patents
Medical image zero watermark generation algorithm based on multi-algorithm fusion Download PDFInfo
- Publication number
- CN114742687A CN114742687A CN202210219858.1A CN202210219858A CN114742687A CN 114742687 A CN114742687 A CN 114742687A CN 202210219858 A CN202210219858 A CN 202210219858A CN 114742687 A CN114742687 A CN 114742687A
- Authority
- CN
- China
- Prior art keywords
- watermark
- medical image
- image
- zero
- original
- 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
- 238000004422 calculation algorithm Methods 0.000 title claims abstract description 98
- 230000004927 fusion Effects 0.000 title claims abstract description 29
- 239000011159 matrix material Substances 0.000 claims abstract description 45
- 238000000354 decomposition reaction Methods 0.000 claims abstract description 23
- 238000000034 method Methods 0.000 claims abstract description 17
- 230000009466 transformation Effects 0.000 claims abstract description 16
- 238000000638 solvent extraction Methods 0.000 claims abstract description 5
- 230000000739 chaotic effect Effects 0.000 claims description 12
- 238000004364 calculation method Methods 0.000 claims description 10
- 238000000605 extraction Methods 0.000 claims description 7
- 238000013507 mapping Methods 0.000 claims description 7
- 238000012545 processing Methods 0.000 claims description 7
- 238000005070 sampling Methods 0.000 claims description 4
- 238000011156 evaluation Methods 0.000 claims description 3
- 230000009467 reduction Effects 0.000 claims description 3
- 238000012935 Averaging Methods 0.000 claims description 2
- 230000000903 blocking effect Effects 0.000 claims description 2
- 238000003672 processing method Methods 0.000 abstract description 2
- 238000013519 translation Methods 0.000 description 13
- 230000009286 beneficial effect Effects 0.000 description 10
- 238000001914 filtration Methods 0.000 description 9
- 238000005516 engineering process Methods 0.000 description 8
- 230000006835 compression Effects 0.000 description 7
- 238000007906 compression Methods 0.000 description 7
- 238000005520 cutting process Methods 0.000 description 6
- 210000004556 brain Anatomy 0.000 description 5
- 230000008859 change Effects 0.000 description 3
- 238000010586 diagram Methods 0.000 description 3
- 238000004091 panning Methods 0.000 description 3
- 230000008569 process Effects 0.000 description 3
- 238000004458 analytical method Methods 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 238000009826 distribution Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000002474 experimental method Methods 0.000 description 2
- 230000006872 improvement Effects 0.000 description 2
- 230000002441 reversible effect Effects 0.000 description 2
- 238000004088 simulation Methods 0.000 description 2
- 238000003860 storage Methods 0.000 description 2
- 230000000007 visual effect Effects 0.000 description 2
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 230000000052 comparative effect Effects 0.000 description 1
- 238000013527 convolutional neural network Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000002059 diagnostic imaging Methods 0.000 description 1
- 230000009977 dual effect Effects 0.000 description 1
- 230000004807 localization Effects 0.000 description 1
- 238000004377 microelectronic Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012552 review Methods 0.000 description 1
- 230000001131 transforming effect Effects 0.000 description 1
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 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
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G16H30/00—ICT specially adapted for the handling or processing of medical images
Landscapes
- Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Radiology & Medical Imaging (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
- Physics & Mathematics (AREA)
- Epidemiology (AREA)
- General Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Primary Health Care (AREA)
- Public Health (AREA)
- Medical Treatment And Welfare Office Work (AREA)
Abstract
The invention relates to the field of image data processing methods, in particular to a medical image zero-watermark generation algorithm based on multi-algorithm fusion, which comprises the following steps: acquiring an original medical image; carrying out non-subsampled contourlet transformation on an original medical image to obtain low-frequency domain information, partitioning the low-frequency domain information, obtaining a coefficient matrix of a low-frequency sub-band by utilizing multi-level discrete cosine transformation, and then constructing a feature vector by using a bivariate decomposition method to serve as a feature matrix for extracting medical features from the original medical image; acquiring an original watermark image, and encrypting the original watermark image; and generating a zero watermark by the encrypted original watermark image and the feature matrix. The method has stronger robustness to common attacks and geometric attacks.
Description
Technical Field
The invention relates to the field of image data processing methods, in particular to a medical image zero-watermark generation algorithm based on multi-algorithm fusion.
Background
With the rapid development of the internet and medical imaging technology, the digitization technology is widely permeating into the field of medical images, and the development of big data and cloud computing brings great convenience to the storage and transmission of medical images in the cloud. And also presents a great challenge to the protection of the relevant information of the patient and the copyright of the medical image. The digital watermarking technology is an information hiding technology, and the problem can be effectively solved by utilizing the digital watermarking technology.
Common digital watermarking algorithms for distortion-free medical images fall into two categories. The first type is reversible watermarking, which is to hide the watermark into the original medical image, and after watermark embedding, the hidden watermark information can be successfully extracted from the medical image, but the robustness is poor. The second type is zero watermarking, which is the information that constructs the watermark using important features of medical images, thus realizing zero watermarking. The zero-watermark algorithm well solves the contradiction between the perceptibility and the robustness of the invisible digital watermark.
At present, a method for processing a medical image by using a zero watermark algorithm includes: based on a robust zero-watermark algorithm of extremely complex exponential transformation and logical mapping, the robustness of the algorithm against geometric attack is improved by utilizing the geometric invariance of PCET; based on a robust zero watermark algorithm of the CNN, the inherent characteristics of each medical image can be quickly extracted by utilizing neural network learning, so that the performance of the algorithm is improved; the medical image robust zero watermark algorithm based on Contourlet transformation and DCT adopts Contourlet transformation to realize decomposition in any direction on any scale, can extract multi-directional image characteristics, and has good practicability in medical and other related fields; the zero-watermark algorithm based on Curvelet-DWT-SVD, combines Curvelet transform and wavelet transform to characterize image characteristics and utilizes the advantages of secondary block subdivision to overcome the defects of DWT-SVD, and has better resistance performance and stability to common geometric attacks; the DTCW-DCT-based medical image zero watermark algorithm combines the visual feature vector of the medical image with encryption technology and third-party concept, solves the problem of fast embedding and watermark extraction, and has strong robustness to conventional attacks and geometric attacks.
Although these algorithms can achieve a certain processing effect, the robustness of these algorithms to geometric attacks such as rotation and translation is poor.
Disclosure of Invention
The invention aims to provide a medical image zero-watermark generation algorithm based on multi-algorithm fusion so as to solve the problem that the existing algorithm is poor in robustness to geometric attacks such as rotation and translation.
The medical image zero-watermark generation algorithm based on multi-algorithm fusion in the scheme comprises the following steps:
step 3, acquiring an original watermark image, and encrypting the original watermark image;
and 4, generating a zero watermark by the encrypted original watermark image and the feature matrix.
The beneficial effect of this scheme is:
the method comprises the steps of carrying out non-downsampling contourlet transformation on an original medical image, extracting low-frequency domain information, carrying out multi-level discrete cosine transformation on the low-frequency domain information, obtaining a coefficient matrix, carrying out biqiyuan decomposition and decomposition to construct a feature vector of the original medical image, and generating a zero watermark by the feature vector and an encrypted original watermark image. The security of original watermark information can be ensured through encryption processing, the integrity of a medical image is ensured by adopting a zero watermark technology when the watermark is embedded, the watermark information can be effectively extracted, the invisibility is good, and the robustness to common attacks and geometric attacks is strong.
Further, in the step 2, the original medical image is subjected to two-level non-down sampling contourlet transformation to obtain a low-frequency sub-band LL2As low frequency domain information LL2Is of a size of
The beneficial effects are that: the low-frequency sub-band is obtained by transforming the medical image through the two-level non-down sampling contourlet, the multi-resolution characteristic and the time-frequency localization analysis capability of the traditional wavelet transform can be kept, and the method has the characteristics of better direction selectivity, translation invariance, limited data redundancy and high-efficiency calculation efficiency, thereby being beneficial to improving the robustness of the watermark.
Further, in the step 2, the step of performing blocking and discrete cosine transform on the low frequency domain information is to perform block division and discrete cosine transform on the LL2Performing a discrete cosine transform, and taking the block of the coefficient matrix at the top left corner of the coefficient matrix as DI1，DI1Is of the size of
Then DI is added1Dividing into non-overlapping sub-blocks of 4 × 4 size, each sub-block being denoted as BrR is 1,2, …, N; for each sub-block BrPerforming a second discrete cosine transform to obtain DBrR is 1,2, …, N; for each sub-block DBrAnd performing ZigZag sorting, and taking the first 9 coefficients to recombine into a block matrix with the size of 3 multiplied by 3.
The beneficial effects are that: the low-frequency domain information of the medical image is processed through discrete cosine transform, the energy compressibility and the decorrelation of the medical image can be well performed, and the energy concentration characteristic of DCT is fully utilized through multi-stage discrete cosine transform, so that more data with larger numerical values can be obtained, the performance of the algorithm can be optimized through the multi-stage DCT, and the algorithm has the capability of fast and accurate feature extraction.
Further, in step 2, the biqiqi decomposition includes performing bidiagonalization decomposition on each block matrix according to a first preset formula to obtain a bidiagonalization matrix δ of each sub-blockrR 1,2, …, N, for each subblock δ according to a first predetermined formularSingular value decomposition is carried out, each singular value matrix is averaged to obtain the average singular value of each subblock, and a block singular value mean value matrix is formed
r is 1,2, …, N, and the first predetermined formula is:
calculating to obtain an integral singular value mean value S according to a second preset formulaavgThe second preset formula is as follows:
the beneficial effects are that: the feature vector is constructed by carrying out bivariate value decomposition on the coefficient matrix of the low-frequency sub-band, so that the accuracy of extracting the specific features from the medical image can be improved, and meanwhile, the safety of the algorithm and the robustness of resisting geometric attacks are also improved.
Further, in the step 2, the singular value mean of the block is compared
With the mean value S of the global singular valuesavgGenerates a binary feature vector T, and the calculation formula is as follows:
the beneficial effects are that: by generating the binary feature vector, the data can be normalized and the amount of data can be reduced.
Further, in the step 3, the original watermark image W is subjected to binarization processing to obtain a binary watermark image, then a chaotic sequence is generated through Logistic mapping, and Logistic chaotic position scrambling is performed on the binary watermark image to obtain a scrambled binary watermark image W1With the scrambled key being K1。
The beneficial effects are that: by carrying out binarization processing on the original watermark image, redundant information is removed, and the watermark is scrambled by using the chaotic sequence, the relevance among pixels can be eliminated, and the security of the watermark is improved.
Further, in the step 4, the scrambled binary watermark image W is processed1And binary feature vector TnPerforming exclusive-or operation on the rows to generate a zero watermark, wherein the expression of the exclusive-or operation is as follows: z ═ XOR (W)1,T)。
The beneficial effects are that: and adding the unique characteristic vector extracted from the medical image into the encrypted watermark image in an exclusive-or operation mode to generate a zero watermark, so that the safety is improved.
Further, the method also comprises a step 5 of carrying out common attack on the medical image, and repeating the step 2 on the attacked medical image to obtain a binary feature vector T'; performing XOR operation on the zero watermark Z and the binary eigenvector T' obtained in the step 4 to extract scrambled watermark information W1', the calculation is: w1'-XOR (Z, T'); the obtained scrambled watermark information W1According to a secret key K1And (3) carrying out logistic chaotic position reduction to obtain decrypted watermark information W', carrying out similarity evaluation on the decrypted watermark information and the zero watermark obtained in the step (4) by adopting a normalized correlation coefficient, wherein when the normalized correlation coefficient is more than 0.9, the extracted watermark information is similar to the zero watermark obtained in the step (4), the damage degree of the attacked medical image is evaluated by adopting a peak signal-to-noise ratio, and when the peak signal-to-noise ratio is more than 10%, the extraction of the zero watermark under the common attack is effective.
The beneficial effects are that: after the zero watermark is generated, the effectiveness of the zero watermark is ensured by applying common attack and extracting watermark information.
Drawings
FIG. 1 is a schematic diagram of a first embodiment of a medical image zero-watermark generation algorithm based on multi-algorithm fusion according to the present invention;
FIG. 2 is a structural diagram of a non-downsampling contourlet transform in an embodiment of a medical image zero-watermark generation algorithm based on multi-algorithm fusion according to the present invention;
FIG. 3 is a DCT coefficient frequency distribution diagram in the first embodiment of the medical image zero-watermark generation algorithm based on multi-algorithm fusion according to the present invention;
fig. 4 is a schematic fierce of watermark extraction in the first embodiment of the medical image zero-watermark generation algorithm based on multi-algorithm fusion;
fig. 5(a) is an original medical image before zero watermark is added in the second embodiment of the medical image zero watermark generation algorithm based on multi-algorithm fusion according to the present invention;
fig. 5(b) is a binary watermark image before adding a zero watermark in the second embodiment of the medical image zero watermark generation algorithm based on multi-algorithm fusion according to the present invention;
fig. 5(c) is a watermark image encrypted before adding a zero watermark in the second embodiment of the medical image zero watermark generation algorithm based on multi-algorithm fusion according to the present invention;
fig. 6(a) is a zero watermark image in a second embodiment of the medical image zero watermark generation algorithm based on multi-algorithm fusion according to the present invention;
fig. 6(b) is a watermark image extracted in the second embodiment of the medical image zero-watermark generation algorithm based on multi-algorithm fusion according to the present invention;
FIG. 7 is a medical image and an extracted watermark image under a normal attack in a second embodiment of a medical image zero-watermark generation algorithm based on multi-algorithm fusion according to the present invention;
FIG. 8 is a medical image and an extracted watermark image under geometric attack in a second embodiment of a medical image zero-watermark generation algorithm based on multi-algorithm fusion according to the present invention;
FIG. 9 is an NC histogram of different methods under normal attack in a second embodiment of the multi-algorithm fusion-based medical image zero-watermark generation algorithm of the invention;
fig. 10 is an NC histogram of different methods under geometric attack in the second embodiment of the multi-algorithm fusion-based medical image zero watermark generation algorithm of the present invention.
Detailed Description
The following is a more detailed description of the present invention by way of specific embodiments.
Example one
The medical image zero watermark generation algorithm based on multi-algorithm fusion, as shown in fig. 1, includes the following steps:
The method comprises the following steps of partitioning low-frequency domain information, acquiring a coefficient matrix of a low-frequency subband by utilizing multi-level Discrete Cosine Transform (DCT), avoiding complex operation and having higher calculation speed, and performing the partitioning and the discrete cosine transform on the low-frequency domain information:
to LL2Performing a discrete cosine transform, and taking the block of the coefficient matrix at the top left corner of the coefficient matrix as DI1，DI1Is of a size of
Then DI is added1Divided into non-overlapping sub-blocks of size 4 x 4, each denoted as BrR is 1,2, …, N; for each sub-block BrPerforming a second discrete cosine transform to obtain DBrR is 1,2, …, N; for each sub-block DBrAnd performing ZigZag sorting, and taking the first 9 coefficients to recombine into a block matrix with the size of 3 multiplied by 3.
The principle of discrete cosine transform is:
for a matrix f (x, y) of size N x N. Where F (x, y) represents an image and F (u, v) represents a corresponding DCT coefficient, the formulas of the two-dimensional forward discrete cosine transform (2D-DCT) and the two-dimensional inverse discrete cosine transform (2D-IDCT) are as follows:
After the image is subjected to discrete cosine transform, a frequency coefficient matrix is obtained, and the distribution of the frequency coefficients is shown in fig. 3. Wherein the upper left corner is the direct current coefficient (DC), and the coefficients are characterized by the maximum amplitude and the minimum frequency; taking the dc coefficient as the origin of coordinates, the values of the coefficients in each direction tend to decrease and the frequencies tend to increase as they extend to the right and downward, these areas being collectively referred to as AC Coefficients (AC). Most of the information of the image is concentrated on the dc coefficient and its low frequency coefficient, and the region farther from the dc coefficient contains less image information. The discrete cosine transform has good energy compressibility and decorrelation.
And then constructing a feature vector by using a bisingular value decomposition method as a feature matrix for extracting medical features from the original medical image, wherein the bisingular value decomposition process comprises the following steps: bibibibisingular value decomposition first performs a bidiagonalization operation on the image, e.g. on the matrix Am×nCarrying out double diagonalization decomposition to obtain a double diagonal matrix as shown in the following formula:
wherein U is1Is an orthogonal matrix, V1 TIs a unitary matrix, δ1Is a double diagonal matrix.
Then, singular value decomposition is carried out on the dual diagonal matrix to obtain a first preset formula as shown in the following formula:
the dual-singular value decomposition not only improves the security of the watermark, but also can describe the characteristic information of the image by using fewer coefficients, thereby constructing the zero watermark with better robustness.
The bibibiqiv value decomposition comprises performing bidiagonalization decomposition on each block matrix according to a first preset formula to obtain a bidiagonalization matrix delta of each sub-blockrR 1,2, …, N, for each subblock δ according to a first predetermined formularPerforming singular value decomposition, averaging each singular value matrix to obtain average singular value of each sub-block, and forming block singular value mean matrix
r＝1,2,…,N；
Calculating to obtain an integral singular value mean value S according to a second preset formulaavgThe second preset formula is as follows:
by comparing block singular value means
With global singular value mean SavgGenerating a binary feature vector T, wherein the calculation formula is that the binary feature vector is used as a feature vector:
and 3, acquiring an original watermark image and encrypting the original watermark image, specifically, performing binarization processing on the original watermark image W to obtain a binary watermark image, generating a chaotic sequence through Logistic mapping, and performing Logistic (Logistic) mapping chaotic position scrambling on the binary watermark image, wherein the scrambled key is K1And randomly scrambling according to the secret key, wherein the scrambling technology is the prior art and is not described herein any more, and the scrambled binary watermark image W is obtained1。
And 4, generating a zero watermark by the encrypted original watermark image and the feature matrix for storage, namely: the scrambled binary watermark image W1And performing XOR operation on the binary characteristic vector T to generate a zero watermark, wherein the XOR operation expression is as follows: Z-XOR (W)1,T)。
And 5, after the zero watermark is generated, extracting the watermark of the medical image added with the zero watermark, wherein the process of extracting the watermark is the reverse process of generating the watermark, the watermark extraction algorithm needs to extract the zero watermark from the stored information, and then the watermark information is extracted by combining the feature matrix extracted from the attacked medical image. I' is the medical image after being attacked, and the size is M multiplied by M. The watermark extraction algorithm, as shown in fig. 4, specifically includes the following steps:
(1) and (3) repeating the step (2) on the attacked medical image I 'to obtain a binary feature vector T'.
(2) Carrying out XOR operation on the zero watermark Z and the binary eigenvector T' obtained in the step 4 to extract scrambled watermark information W1'. The calculation formula is as follows:
W1′＝XOR(Z,T′)。
(3) the obtained scrambled watermark information W1According to a secret key K1And performing Logistic chaotic position reduction to obtain decrypted watermark information W'.
Example two
The difference between the medical image zero-watermark generation algorithm based on multi-algorithm fusion and the first embodiment is that in order to verify the effectiveness of the first embodiment, simulation experiments are carried out on a 64-bit Windows 10 operating system and a MATLAB R2019b platform.
In the second embodiment, the tenth slice of the MRI brain map is selected from a large amount of medical data as the original medical image, and the size is 128 × 128 pixels, as shown in fig. 5 (a). The original watermark image is an effective binary image of 64 × 64 pixels, as shown in fig. 5 (b). The key used in Logistic chaotic scrambling is x0Fig. 5(c) shows the watermark after Logistic chaotic encryption, where 0.6 and μ 4. As is clear from the figure, the watermark image is changed in a way that can not be identified by naked eyes, which improves the security of the watermark. Fig. 6(a) is a zero watermark image, the image remains unchanged both before and after watermark embedding, and the watermarked medical image remains identical to the original image when not under attack, with the extracted corresponding watermark having an NC value of 1.00, as shown in fig. 5 (b).
The algorithm of the first embodiment is evaluated through robustness, the similarity degree between the extracted watermark image and the original watermark image is evaluated by adopting a normalized correlation coefficient (NC) as an evaluation standard, and the damage degree of the attacked medical image is evaluated by adopting a peak signal-to-noise ratio (PSNR). The calculation formulas of NC and PSNR are as follows:
wherein, W (i, j) is the pixel value of the original watermark image, W' (i, j) is the pixel value of the extracted watermark image, N is the size of the watermark image, and MSE is the mean square error of the original medical image and the attacked medical image.
Common attacks are performed on medical images, and the common attacks include noise attacks, compression attacks, and filtering attacks. Fig. 7 shows an image of a medical image after a general attack and an extracted watermark image, wherein: (a) medical images under Gaussian noise attack with a mean value of 0 and a variance of 0.01; (b) the mean value is 0, and the variance is 0.01; (c) medical images under Gaussian noise attack with a mean value of 0 and a variance of 0.3; (d) the mean value is 0, and the variance is 0.3; (e) compressing the medical image under JPEG compression attack with the mass of 2%; (f) compressing the extracted watermark image under the JPEG compression attack with the mass of 2%; (g) the filtering window is [7x7], and the medical image under the 20time median filtering attack is obtained; (h) the filtering window is [7x7], 20 meters of median filtering attack.
As can be seen from fig. 7, the watermark image extracted from the medical image after being subjected to a common attack is still clear, and particularly under the conditions of gaussian noise attack and median filtering attack, the medical image has great changes in shape and detail, and is hardly recognizable compared with the original medical image, but the extracted watermark image still has a good visual effect.
In order to further analyze the robustness of the algorithm under common attacks with different strengths, the medical image is subjected to common attacks with different degrees, and the experimental results are shown in table 1.
TABLE 1 NC and PSNR values of MRI brain plots under general attack
As can be seen from the data in table 1, for the gaussian noise attack, as the noise intensity increases, the NC value changes only slightly and is kept above 0.98; for the JPEG compression attack, even under the condition that the compression quality is 2%, the NC value can still reach 0.98, and when the compression quality is more than 5%, the obtained NC values are all 1; for the median filtering attack, under the condition that the size of the template is 7 multiplied by 7 and the repetition is carried out for 20times, the NC value can still reach 0.95, and the watermark information can still be extracted more completely. Experiments show that the algorithm has better robustness under common attacks with different strengths. Because the DTCTT transformation has good denoising performance and the obtained low-frequency sub-band has strong stability, the algorithm in the text shows strong robustness in resisting common attacks.
The method comprises the following steps of carrying out geometric attack on a medical image, wherein the geometric attack refers to the unrecoverable change of the image, and common geometric attacks comprise a zooming attack, a rotating attack, a translating attack and a cropping attack. Fig. 8 is an image of a medical image after geometric attack and an extracted watermark image, where: (a) medical images rotated 30 ° clockwise; (b) watermark images extracted under the attack of clockwise rotation of 30 degrees; (c) medical images reduced by 0.125 times and then enlarged by 8 times; (d) reducing the watermark image extracted under the attack of 0.125 times and then amplifying the watermark image by 8 times; (e) medical images translated 25% to the left; (f) watermark images extracted under the attack of leftward translation by 25%; (g) medical images translated down 25%; (h) translating the watermark image extracted under the attack of 25% downwards; (i) cutting 30% of medical images in the X-axis direction; (j) cutting 30% of watermark images extracted under the attack in the X-axis direction; (k) cutting 25% of medical images and the extracted watermark images in the Y-axis direction; (l) Watermark image extracted under attack of cutting 25% in Y-axis direction
As can be seen from fig. 9, the medical image after the geometric attack is severely distorted, and especially under the conditions of the cropping attack and the translation attack, most of information of the medical image is lost and cannot appear in the original form, but the extracted watermark image is still clear and easy to recognize, and the NC value is kept above 0.9.
The results of the experiment are shown in table 2.
Table 2 NC values and PSNR values of MRI brain maps under geometric attacks
As can be seen from the data in table 2, for the scaling attack, the obtained NC values are all 1 under the general scaling attack, and the NC values are all maintained above 0.95 under the high-intensity scaling condition; for a rotational attack, NC remains above 0.89 with a clockwise rotation of 40 °; for a panning attack, the NC value may still reach above 0.92 when panning left by 23% or panning down by 25%. Since the rotation attack and the translation attack change the overall position of the image pixel, and cause the situation of feature shortage, the NC value is lower compared with the non-geometric attack, but the overall NC value still remains about 0.9. The main reason is that the DTCTWT and bisingularity decomposition are used in the algorithm, so that the constructed feature matrix has translation invariance and rotation invariance, the change under the geometric attack is small, meanwhile, the area in direct contact with the medical image is effectively reduced by a sub-block mapping mode, and the robustness of the algorithm against the rotation attack and the translation attack is improved to a certain extent. For the clipping attack, the NC value remains around 0.95 for 25% of clipping. This is because the cropping attack only changes the pixel value size of the cropping position, and only a small part of the features are affected, so the algorithm can resist the cropping attack at different positions. The algorithm herein exhibits a strong robustness against geometric attacks.
Comparing the Algorithm of example one with the existing method to verify the robustness of example one, selecting the 10 th slice of MRI brain Medical image of size 128 × 128 as the original Medical image, the watermark image selecting the binary image of size 64 × 64 with document [1], Jiangzao, Chenmicro, DWT-DCT-SVD Based color image zero Watermarking Algorithm [ J ]. microelectronics and Computer, 2016,33(08): 107. application for horizontal Watermarking in recording works status [ J ]. Computer, plain memory, and Rob application, Digital Watermarking [ 27 (2018) ], document [15] Liu J, Li, Zhang K, Zhang-Water computing ] surface area view [ J ]. Computer, computing information Review, 45(2018), document [15] Liu J, Zhang K, Zhang-computing, and computing, simulation for computing and computing, 2019,9: 188-; (b) the JPEG is compressed; (c) for median filtering, (a) is clockwise rotation in fig. 10; (b) zooming; (c) left translation; (d) is a lower translation; (e) cutting for an X axis; (f) and cutting for the Y axis.
Table 3 comparative experimental data table (NC values) of algorithm and literature under different attacks
As can be seen from table 3, the algorithm herein exhibits stronger robustness against common attacks and geometric attacks than documents [1], [12], [15] and [28 ]. For gaussian noise attack and compression attack, it can be seen from fig. 9 that the NC values obtained by the five algorithms are almost the same, but the algorithm herein still has a slight improvement. For median filtering attack, the algorithm is slightly lower than that of the document [28], but the algorithm is promoted compared with the documents [1], [12] and [15 ]. Among them, the robustness of the document [1] is relatively poor because the wavelet transform used in the document [1] has no directivity, and therefore some of the contour information of the image is lost when a general attack is applied. The algorithm and the Contourlet transform, DTCTWT transform and Curvelet transform used in the document [12], the document [15] and the document [28] have direction selectivity, effectively enhance the outline information of the image, and therefore the image is more robust against common attacks.
For geometric attacks, the algorithm herein has a greater degree of improvement compared to documents [1], [12], [15] and [28], especially in terms of rotational attacks, translational attacks and clipping attacks. As can be seen from fig. 10, the NC value of the algorithm is higher than that of documents [1], [12], [15] and [28] for different angles of rotation attack, and when the rotation is 15% clockwise, the NC value of the algorithm is still about 0.93, and the NC values of documents [1], [12] and [15] are all lower than 0.9, and the robustness is poor. For the translation attack and the clipping attack, the NC value of the algorithm is improved by about 5% -20% compared with the NC value of the algorithm in the documents [1], [12], [15] and [28 ].
The foregoing is merely an example of the present invention and common general knowledge of known specific structures and features of the embodiments is not described herein in any greater detail. It should be noted that, for those skilled in the art, without departing from the structure of the present invention, several changes and modifications can be made, which should also be regarded as the protection scope of the present invention, and these will not affect the effect of the implementation of the present invention and the practicability of the patent. The scope of the claims of the present application shall be determined by the contents of the claims, and the description of the embodiments and the like in the specification shall be used to explain the contents of the claims.
Claims (8)
1. The medical image zero-watermark generation algorithm based on multi-algorithm fusion is characterized by comprising the following steps of:
step 1, acquiring an original medical image;
step 2, carrying out non-downsampling contourlet transformation on the original medical image to obtain low-frequency domain information, partitioning the low-frequency domain information, obtaining a coefficient matrix of a low-frequency sub-band by utilizing multi-level discrete cosine transformation, and then constructing a feature vector by using a bivariate decomposition method to serve as a feature matrix for extracting medical features from the original medical image;
step 3, acquiring an original watermark image, and encrypting the original watermark image;
and 4, generating a zero watermark by the encrypted original watermark image and the feature matrix.
2. The medical image zero-watermark generation algorithm based on multi-algorithm fusion according to claim 1, characterized in that: in the step 2, the original medical image is subjected to two-stage non-down sampling contourlet transformation to obtain a low-frequency sub-band LL2As low frequency domain information LL2Is of the size of
3. The medical image zero-watermark generation algorithm based on multi-algorithm fusion as claimed in claim 2, wherein: in the step 2, the step of carrying out blocking and discrete cosine transform on the low-frequency domain information comprises the step of carrying out block division and discrete cosine transform on LL2Performing a discrete cosine transform, and taking the block of the coefficient matrix at the top left corner of the coefficient matrix as DI1，DI1Is of the size of
Then DI is added1Divided into non-overlapping sub-blocks of size 4 x 4, each denoted as BrR is 1,2, …, N; for each sub-block BrPerforming a second discrete cosine transform to obtain DBrR is 1,2, …, N; for each sub-block DBrAnd performing ZigZag sorting, and taking the first 9 coefficients to recombine into a block matrix with the size of 3 multiplied by 3.
4. The medical image zero-watermark generation algorithm based on multi-algorithm fusion according to claim 3, characterized in that: in the step 2, the biqiqi decomposition includes performing bidiagonalization decomposition on each block matrix according to a first preset formula to obtain a bidiagonalization matrix δ of each sub-blockrR 1,2, …, N, for each subblock δ according to a first predetermined formularPerforming singular value decomposition, averaging each singular value matrix to obtain average singular value of each sub-block, and forming block singular value mean matrix
The first predetermined formula is:
calculating to obtain an integral singular value mean value S according to a second preset formulaavgThe second preset formula is as follows:
5. the medical image zero-watermark generation algorithm based on multi-algorithm fusion according to claim 4, wherein: in the step 2, the singular value mean of the block is compared
With global singular value mean SavgGenerates a binary feature vector T, and the calculation formula is as follows:
6. the medical image zero-watermark generation algorithm based on multi-algorithm fusion according to claim 5, wherein: in the step 3, the original watermark image W is subjected to binarization processing to obtain a binary watermark image, then a chaotic sequence is generated through logistic mapping, and the binary watermark image is subjected to logistic mapping chaotic position scrambling to obtain a scrambled binary watermark image W1The scrambled key is K1。
7. The medical image zero-watermark generation algorithm based on multi-algorithm fusion as claimed in claim 6, wherein: in the step 4, the scrambled binary watermark image W is processed1And performing XOR operation on the binary characteristic vector T to generate a zero watermark, wherein the XOR operation expression is as follows: z ═ XOR (W)1,T)。
8. The medical image zero-watermark generation algorithm based on multi-algorithm fusion of claim 6, wherein: further comprising a step 5 of performing on the medical imageIn the general attack, repeating the step 2 on the attacked medical image to obtain a binary feature vector T'; performing XOR operation on the zero watermark Z and the binary eigenvector T' obtained in the step 4 to extract scrambled watermark information W1', the calculation is: w1'-XOR (Z, T'); the obtained scrambled watermark information W1According to a secret key K1And (3) carrying out logistic chaotic position reduction to obtain decrypted watermark information W', carrying out similarity evaluation on the decrypted watermark information and the zero watermark obtained in the step (4) by adopting a normalized correlation coefficient, wherein when the normalized correlation coefficient is more than 0.9, the extracted watermark information is similar to the zero watermark obtained in the step (4), the damage degree of the attacked medical image is evaluated by adopting a peak signal-to-noise ratio, and when the peak signal-to-noise ratio is more than 10%, the extraction of the zero watermark under the common attack is effective.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210219858.1A CN114742687A (en) | 2022-03-08 | 2022-03-08 | Medical image zero watermark generation algorithm based on multi-algorithm fusion |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210219858.1A CN114742687A (en) | 2022-03-08 | 2022-03-08 | Medical image zero watermark generation algorithm based on multi-algorithm fusion |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114742687A true CN114742687A (en) | 2022-07-12 |
Family
ID=82274480
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210219858.1A Pending CN114742687A (en) | 2022-03-08 | 2022-03-08 | Medical image zero watermark generation algorithm based on multi-algorithm fusion |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114742687A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115311119A (en) * | 2022-10-09 | 2022-11-08 | 中国民航大学 | Three-dimensional image zero watermark embedding and extracting method capable of resisting geometric attack |
| CN116883226A (en) * | 2023-07-21 | 2023-10-13 | 中国国土勘测规划院 | NMF decomposition-based DEM zero watermark method, device and medium |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6975733B1 (en) * | 1999-09-10 | 2005-12-13 | Markany, Inc. | Watermarking of digital images using wavelet and discrete cosine transforms |
| CN103955879A (en) * | 2014-04-11 | 2014-07-30 | 杭州电子科技大学 | DWT-SVD robust blind watermark method based on multilevel DCT |
| CN111968025A (en) * | 2020-08-19 | 2020-11-20 | 海南大学 | Bandlelet-DCT-based medical image robust zero watermarking method |
| AU2020104204A4 (en) * | 2020-12-21 | 2021-03-11 | South China University Of Technology | A method and a device for generating a digital watermark image based on graphic codes |
-
2022
- 2022-03-08 CN CN202210219858.1A patent/CN114742687A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6975733B1 (en) * | 1999-09-10 | 2005-12-13 | Markany, Inc. | Watermarking of digital images using wavelet and discrete cosine transforms |
| CN103955879A (en) * | 2014-04-11 | 2014-07-30 | 杭州电子科技大学 | DWT-SVD robust blind watermark method based on multilevel DCT |
| CN111968025A (en) * | 2020-08-19 | 2020-11-20 | 海南大学 | Bandlelet-DCT-based medical image robust zero watermarking method |
| AU2020104204A4 (en) * | 2020-12-21 | 2021-03-11 | South China University Of Technology | A method and a device for generating a digital watermark image based on graphic codes |
Non-Patent Citations (11)
| Title |
|---|
| T. HUANG, J. XU, Y. YANG, S. TU AND B. HAN: "Zero-Watermarking Algorithm for Medical Images Based on Nonsubsampled Contourlet Transform and Double Singular Value Decomposition", 《2021 5TH ASIAN CONFERENCE ON ARTIFICIAL INTELLIGENCE TECHNOLOGY (ACAIT)》, pages 65 - 76 * |
| 刘万军;孙思宇;冯琳;姜文涛: "双树复小波域的稳健性数字水印算法", 《辽宁工程技术大学学报(自然科学版)》, vol. 37, no. 3, pages 660 - 668 * |
| 刘万军;孙思宇;冯琳;姜文涛;: "双树复小波域的稳健性数字水印算法", 辽宁工程技术大学学报(自然科学版), no. 03, pages 660 - 668 * |
| 展虎;毛力: "基于NSCT-DCT-DWT-SVD 联合数字水印算法", 《计算机应用与软件》, vol. 35, no. 3, pages 309 - 315 * |
| 展虎;毛力;: "基于NSCT-DCT-DWT-SVD联合数字水印算法", 计算机应用与软件, no. 03, pages 309 - 315 * |
| 暴琳;李垣江;龚淼;张贞凯;: "基于Tetrolet变换和DCT-SVD的彩色图像稳健数字水印算法", 江苏科技大学学报(自然科学版), no. 05, pages 454 - 461 * |
| 王琳娟;李君霞;: "基于分块的多级DCT和SVD的鲁棒水印算法", 计算机时代, no. 09, pages 51 - 55 * |
| 肖振久;张晗;陈虹;高婷;: "增强奇异值分解和细胞神经网络的零水印", 中国图象图形学报, no. 03, pages 288 - 296 * |
| 金炜;励金祥;尹曹谦;: "一种基于可视密码的轮廓小波图像零水印方案", 光电子.激光, no. 05, pages 653 - 656 * |
| 齐向明;张晶;谭昕奇: "基于低频奇异值均值的强鲁棒零水印算法", 《计算机工程》, vol. 45, no. 12, pages 214 - 221 * |
| 齐向明;张晶;谭昕奇;: "基于低频奇异值均值的强鲁棒零水印算法", 计算机工程, no. 12, pages 214 - 221 * |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115311119A (en) * | 2022-10-09 | 2022-11-08 | 中国民航大学 | Three-dimensional image zero watermark embedding and extracting method capable of resisting geometric attack |
| CN115311119B (en) * | 2022-10-09 | 2022-12-23 | 中国民航大学 | Three-dimensional image zero watermark embedding and extracting method capable of resisting geometric attack |
| CN116883226A (en) * | 2023-07-21 | 2023-10-13 | 中国国土勘测规划院 | NMF decomposition-based DEM zero watermark method, device and medium |
| CN116883226B (en) * | 2023-07-21 | 2024-01-02 | 中国国土勘测规划院 | DEM zero watermark embedding and extracting method, device and medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Ahmaderaghi et al. | Blind image watermark detection algorithm based on discrete shearlet transform using statistical decision theory | |
| Evsutin et al. | Watermarking schemes for digital images: Robustness overview | |
| Andalibi et al. | Digital image watermarking via adaptive logo texturization | |
| Li et al. | Dither modulation of significant amplitude difference for wavelet based robust watermarking | |
| Assini et al. | A robust hybrid watermarking technique for securing medical image | |
| CN111968025A (en) | Bandlelet-DCT-based medical image robust zero watermarking method | |
| Wang et al. | A novel zero-watermarking scheme based on variable parameter chaotic mapping in NSPD-DCT domain | |
| CN112381701B (en) | RST attack resisting stereo image zero watermarking method based on FFST and Hessenberg decomposition | |
| CN114742687A (en) | Medical image zero watermark generation algorithm based on multi-algorithm fusion | |
| Zhang et al. | Image watermarking scheme based on Arnold transform and DWT-DCT-SVD | |
| Chen et al. | A robust watermarking algorithm based on QR factorization and DCT using quantization index modulation technique | |
| Gao et al. | A digital watermarking algorithm for color image based on DWT | |
| Kekre et al. | Performance comparison of DCT and walsh transforms for watermarking using DWT-SVD | |
| Rahman et al. | DWT-SVD based new watermarking idea in RGB color space | |
| AlShaikh et al. | A novel CT scan images watermarking scheme in DWT transform coefficients | |
| Gaur et al. | A RDWT and block-SVD based dual watermarking scheme for digital images | |
| Liu et al. | Optimal blind watermarking for color images based on the U matrix of quaternion singular value decomposition | |
| Ayangar et al. | A novel DWT-SVD based watermarking scheme | |
| Hoshi et al. | A robust watermark algorithm for copyright protection by using 5-level DWT and two logos | |
| Kejgir et al. | Lifting wavelet transform with singular value decomposition for robust digital image watermarking | |
| Bhuyan et al. | Shuffled SVD based robust and secure digital image watermarking | |
| Gupta et al. | Improved RST-attacks resilient image watermarking based on joint SVD-DCT | |
| Verma et al. | Watermarking scheme based on singular value decomposition and homomorphic transform | |
| Shi | DWT and SVD based Watermarking Scheme with Circulation. | |
| Nesakumari et al. | Normalized image watermarking scheme using chaotic system |
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 |