CN111048185B - Interesting region parameter game analysis method based on machine learning - Google Patents
Interesting region parameter game analysis method based on machine learning Download PDFInfo
- Publication number
- CN111048185B CN111048185B CN201911356216.0A CN201911356216A CN111048185B CN 111048185 B CN111048185 B CN 111048185B CN 201911356216 A CN201911356216 A CN 201911356216A CN 111048185 B CN111048185 B CN 111048185B
- Authority
- CN
- China
- Prior art keywords
- region
- interest
- image
- encryption
- medical 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.)
- Active
Links
Images
Classifications
-
- 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
- G16H30/40—ICT specially adapted for the handling or processing of medical images for processing medical images, e.g. editing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/11—Region-based segmentation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20092—Interactive image processing based on input by user
- G06T2207/20104—Interactive definition of region of interest [ROI]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30004—Biomedical image processing
Abstract
A machine learning-based interesting region parameter game analysis method relates to the technical field of medical image processing and information security, and provides a machine learning-based interesting region parameter game analysis method for solving the problems that the encryption performance of a medical image is affected and the like due to the fact that the dividing of an existing medical image interesting region is not accurate enough. Taking an L-bit gray level medical image with the size of M multiplied by N as an original medical image Pimg, setting the number of game participants as P, and setting ROI of the participants i The method comprises the steps of analyzing the encryption performance of the region of interest according to a global region of interest threshold tau under the constraint condition of a strategy to obtain an encryption effect set, and fitting the encryption effect set SetEnEffect by adopting a machine learning polynomial linear regression method. And providing reference basis for the partition parameter setting of the threshold segmentation algorithm.
Description
Technical Field
The invention relates to the technical field of medical image processing and information security, in particular to a region-of-interest parameter game analysis method based on machine learning.
Background
With the development of the internet and multimedia technology, various information is spread through the internet, and the information security problem is receiving wide attention. The safe storage and transmission of medical images by using computer technology is a main trend for the development of systematic management and modern informatization construction of hospitals. Medical images store very important pathological information of patients, and have higher confidentiality requirements. The digital medical images at present mainly include CT (computed tomography) images, MRI (magnetic resonance imaging) images, B-mode scanning images, and the like. Compared with common digital images, the medical image has larger data volume and higher redundancy, the medical image can be divided into a region of interest (ROI) and a region of non-interest (ROI), in order to realize real-time safety service on the limited equipment, the medical image adopts a partial encryption scheme, namely only the ROI is encrypted, so that not only can the information safety be protected, but also the computing resources can be saved and the encryption speed can be improved. The current region-of-interest extraction algorithm mainly comprises threshold segmentation, edge segmentation, face recognition, infrared segmentation and the like, and most of the schemes are not suitable for various medical image types due to the single region-of-interest division method. Moreover, since the region of interest division directly affects the encryption performance of the medical image, an excessively large or small region of interest directly results in excessively low security of image encryption or excessively slow encryption speed.
Disclosure of Invention
The invention provides a machine learning-based interesting area parameter game analysis method, aiming at solving the problems that the encryption performance of medical images is influenced and the like due to the fact that the dividing of the interesting areas of the existing medical images is not accurate enough. The method for analyzing the parameters of the interesting region in the game based on machine learning is realized by the following steps:
taking an L-bit gray level medical image with the size of M multiplied by N as an original medical image Pimg, wherein L is the pixel depth of the image;
step two, setting the number of game participants as P and the participants as ROI i I =1,2,3, P selects different region of interest partitioning parameters to partition the region of interest, where the participant ROI i The specific method for dividing the region of interest is as follows:
step two, dividing the original medical image Pimg into image blocks with the size of N multiplied by N to obtain M multiplied by N/(N multiplied by N) image blocks
Step two and three, the ROI of the participant i The discrimination method of the region of interest is as follows:
when the temperature is higher than the set temperatureIf so, then the image block is->Setting a corresponding region of interest flag for the region of interest>Is 1; />
When in useIf so, then the image block is->Setting a corresponding region of interest flag for a region of non-interest>Is 0;
where t _ ROI i ROI for participants i Threshold of region of interest, t _ ROI i ∈[0,2 L -1) region of interest flagBlock representation of 1Participant ROI i The region of interest block, namely the selected encryption region block;
step three, setting ROI of the participant i The constraint conditions of the strategy are as follows:
fourthly, performing encryption performance analysis on the region of interest according to a global region of interest threshold tau to obtain an encryption effect set, wherein the global region of interest threshold tau belongs to the 0,2 L -1), the specific analysis steps are:
step four, sequentially encrypting the region-of-interest blocks of the original medical image Pimg by using a global region-of-interest threshold tau to obtain a ciphertext image set SetEnImg;
step two, recording the encryption time of the region-of-interest block for encrypting the original medical image Pimg by using the global region-of-interest threshold tau in sequence to obtain an encrypted image time set SetEnTime;
step four, sequentially calculating a peak signal-to-noise ratio between the region-of-interest block ciphertext image of the original medical image Pimg and the original medical image encrypted by the global region-of-interest threshold tau to obtain a peak signal-to-noise ratio set SetEnPSNR;
step four, sequentially calculating the structural similarity between the region-of-interest block ciphertext image encrypted by the global region-of-interest threshold tau of the original medical image Pimg and the original medical image to obtain a structural similarity set SetEnSSIM;
step four, sequentially calculating the information entropy of the region-of-interest block ciphertext image obtained by encrypting the original medical image Pimg by using the global region-of-interest threshold value to obtain a ciphertext information entropy set SetEntrack;
step four, calculating and obtaining an encryption effect set SetEnEffect according to the encryption image time set SetEnTime, the peak signal-to-noise ratio set SetEnPSNR, the structural similarity set SetEnSSIM and the ciphertext information entropy set SetEncopy described in the step four to the step four; represented by the formula:
SetEnEffect={EnEffect(τ(1)),EnEffect(τ(2)),...,EnEEffect(τ(2 L -1))}
EnEffect(τ)=1/log(ω 1 P 1 EnTime(τ)+ω 2 EnPSNR(τ)+ω 3 P 2 EnSSIM(τ)+ω 4 EnEntropy(τ) (6)
wherein, ω is 1 As a weight of the encryption time, omega 2 As peak signal-to-noise ratio weight, ω 3 As a structural similarity weight, ω 4 Is the information entropy weight; p is 1 For encrypting the time amplification factor, P 2 The amplification factor is the structural similarity;
step five, fitting the encryption effect set SetEnEffect by adopting a machine-learning polynomial linear regression method to obtain an encryption effect gain function SetEnEffect Income;
step six, when the participant ROI i If no strategy is executed, the encryption performance of the steps from four to four is calculated to obtain the initial profit value SetEneffectIncome of the encryption effect profit function SetEneffectIncome 0 (ii) a Set min (SetEneffect') > SetEneffect Income 0 Min () is the function for solving the minimum element value of the set;
step seven, obtaining the initial profit value SetEneffectIncome of the encryption effect profit function SetEneffectIncome according to the step six 0 Obtaining a game Nash bargaining counter-price solution of the area of interest as follows:
obtaining optimal game solution tau of threshold value of region of interest by adopting Lagrange multiplier method optimal ;
Step eight, adopting the area-of-interest threshold game optimal solution tau in the step seven optimal And dividing the interested region of the original medical image Pimg, and encrypting the interested region to obtain an encrypted image.
The invention has the beneficial effects that: according to the interesting region parameter game analysis method based on machine learning, disclosed by the invention, the interesting region division parameters of the medical image are quantitatively analyzed by utilizing the machine learning method and the game theory, so that the encryption safety performance of the medical image is ensured, and the encryption efficiency is also ensured. And providing reference basis for the partition parameter setting of the threshold segmentation algorithm.
Drawings
Fig. 1 is a flowchart of a method for analyzing a region of interest parameter game based on machine learning according to the present invention;
fig. 2 is an encryption effect analysis diagram of the region of interest parameter game analysis method based on machine learning according to the present invention: fig. 2A is a distribution graph of an encrypted image time set along with a threshold value, fig. 2B is a distribution graph of a peak signal-to-noise ratio set along with a threshold value, fig. 2C is a distribution graph of a structure similarity set along with a threshold value, fig. 2D is a distribution graph of a ciphertext information entropy set along with a threshold value, and fig. 2E is a distribution graph of an encryption effect set along with a threshold value;
FIG. 3 is a graph of a machine learning polynomial linear regression fitted encryption effect set gain function;
fig. 4 is an effect diagram of the gaming analysis method for parameters of regions of interest in machine learning according to the present invention: fig. 4A is a "maximum intensity projection of the chest" gray-scale medical image original image, and fig. 4B is a "maximum intensity projection of the chest" gray-scale medical image encrypted image.
Detailed Description
The first embodiment is described with reference to fig. 1 to 3, and the method for analyzing the game of the parameters of the region of interest based on machine learning is implemented by the following steps:
taking an L-bit gray level medical image with the size of M multiplied by N as an original image Pimg, wherein L represents the pixel depth of the image.
Step two, the number of game participants is P, and the participants ROI i I =1,2,3, P selects different region of interest partitioning parameters to partition the region of interest, where the participant ROI i The specific method for dividing the region of interest is as follows:
step two, dividing the original image Pimg into image blocks with the size of nxn to obtain M multiplied by N/(nxn) image blocks
Step two, sequentially calculating the image blocks in the step twoAverage of gray values of all pixels within
Step two and three, participant ROI i The region of interest discrimination method is as follows:
when in useIf so, then the image block is->Is the region of interest. Setting the corresponding region of interest flag->Is 1.
When in useIf so, then the image block is->Is a region of non-interest. Setting the corresponding region of interest flag->Is 0.
Where t _ ROI i ROI for participants i Threshold of region of interest, t _ ROI i ∈[0,2 L -1) region of interest flagBlock 1 indicates participant ROI i I.e. the selected encryption area block.
Step threeParticipant ROI i The constraints of the strategy are:
step four, performing encryption performance analysis on the region of interest according to a global region of interest threshold tau to obtain an encryption effect set, wherein the global region of interest threshold is in a range of tau epsilon [0,2 ] L -1), τ ∈ N, the specific analysis steps are as follows:
step four one, sequentially with a global region of interest threshold τ =0,1,2 L -2 encrypting a region of interest block of the medical image Pimg, resulting in a set of ciphertext images SetEnImg = { EnImg (τ (1)), enImg (τ (2)), · L -1))}。
Step four, sequentially recording the global region of interest threshold value tau =0,1,2 L -2 encrypting the encryption time of the region of interest block of the medical image Pimg resulting in an encrypted image time set:
SetEnTime={EnTime(τ(1)),EnTime(τ(2)),...,EnTime(τ(2 L -1))} (2)
step four and three, calculating a global region of interest threshold value tau =0,1,2 in sequence L -2 encrypting the peak signal-to-noise ratio between the region-of-interest block ciphertext image of the medical image Pimg and the original image to obtain a peak signal-to-noise ratio set:
SetEnPSNR={EnPSNR(τ(1)),EnPSNR(τ(2)),...,EnPSNR(τ(2 L -1))} (3)
step four, calculating a global region of interest threshold value tau =0,1,2 in sequence L -2, encrypting the structural similarity between the region of interest block ciphertext image of the medical image Pimg and the original image to obtain a structural similarity set:
SetEnSSIM={EnSSIM(τ(1)),EnSSIM(τ(2)),...,EnSSIM(τ(2 L -1))} (4)
step four and five, calculating a global region of interest threshold value tau =0,1,2 in sequence L -2 region of interest block density of encrypted medical image PimgAnd (3) obtaining a ciphertext information entropy set by using the information entropy of the text image:
SetEnEntropy={EnEntropy(τ(1)),EnEntropy(τ(2)),...,EnEntropy(τ(2 L -1))} (5)
step four, calculating to obtain an encryption effect set according to the encryption performance sets in the step four, the step two, the step four and the step five:
SetEnEffect={EnEffect(τ(1)),EnEffect(τ(2)),...,EnEEffect(τ(2 L -1))}
EnEffect(τ)=1/log(ω 1 P 1 EnTime(τ)+ω 2 EnPSNR(τ)+ω 3 P 2 EnSSIM(τ)+ω 4 EnEntropy(τ)) (6)
wherein: omega 1 ,ω 2 ,ω 3 ,ω 4 For each encryption performance set weight, omega, from step four two to step four five 1 As a weight of the encryption time, omega 2 As peak signal-to-noise ratio weight, ω 3 As a structural similarity weight, ω 4 Is the information entropy weight; p 1 ,P 2 To enlarge the factor, P 1 For encrypting the time amplification factor, P 2 Is the magnification factor of the structural similarity.
Step five, fitting the encryption effect set SetEnEffect by using a machine-learned polynomial linear regression method to obtain an encryption effect gain function SetEnEffect Income, wherein the specific fitting method is as follows:
and fifthly, amplifying the encryption effect set SetEnEffect in the fourth step in proportion to obtain SetEnEffect' which is used as a result set of polynomial linear regression.
Step five two, X =1,2 L -1 as input set.
And fifthly, increasing the high-order item of X to obtain a high-order item input set Poly _ X.
And fifthly, approximating the result set SetEnEffect' by adopting a polynomial regression method.
And fifthly, obtaining each item coefficient of the polynomial according to the fitting result to obtain an encryption effect gain function SetEneffectIncome.
Step six, when the participant ROI i If no strategy is executed, an encryption scheme of the global image is adopted, the encryption performances of the fourth step, the second step, the fourth step and the fifth step are calculated, and the initial profit value SetEneffiecomen of the encryption effect profit function SetEneffiecomen is obtained 0 . Wherein min (SetEneffect') > SetEneffect Income 0 Min () is the function that finds the minimum element value of the set.
Step seven, the game Nash bargaining price counter-offer solution of the area of interest is expressed as:
obtaining optimal solution tau of threshold game of interest region by using Lagrange multiplier method optimal 。
Step eight, using the region of interest threshold game optimal solution tau in the step seven optimal And dividing the interested region of the original image Pimg, and encrypting the interested region to obtain an encrypted image.
In a second embodiment, the second embodiment is described with reference to fig. 1 to 4, and the method for analyzing the game of the parameters of the region of interest based on machine learning is implemented by the following steps:
step one, taking an L-bit grayscale medical image with a size of 512 × 512 as an original image Pimg, as shown in fig. 4A, where L represents a pixel depth of the image. L =8 in this example.
Step two, the number of game participants is P =1000, and the participants ROI i I =1,2,3, P selects different region of interest partitioning parameters to partition the region of interest, where the participant ROI i The specific method for dividing the region of interest is as follows:
step two, dividing the original image Pimg into image blocks of N × N, where N =8 in this embodiment, to obtain M × N/(N × N) image blocks
Step two, sequentially calculating the image blocks in the step twoAverage of gray values of all pixels within
Step two and three, participant ROI i The region of interest discrimination method is as follows:
when in useIf so, then the image block is->Is a region of interest. Setting the corresponding region of interest flag->Is 1.
When in useIf so, then the image block is->Is a region of non-interest. Setting the corresponding region of interest flag->Is 0.
Where t _ ROI i ROI for participants i Threshold of region of interest, t _ ROI i E [0, 255)), region of interest flag bitBlock 1 indicates participant ROI i I.e. the selected encryption area block.
Step three, participant ROI in this example i The constraints of the strategy are:
fourthly, carrying out region-of-interest encryption performance analysis according to a global region-of-interest threshold value tau to obtain an encryption effect set, wherein the global region-of-interest threshold value range isThe specific analysis steps are as follows:
and step four, sequentially encrypting the region-of-interest blocks of the medical image Pimg by using a global region-of-interest threshold value tau =0,1,2,.. And 254 to obtain a ciphertext image set SetEnImg = { EnImg (tau (1)), enImg (tau (2)),. And EnImg (tau (255))).
And step two, sequentially recording the encryption time of the region-of-interest block of the medical image Pimg encrypted by a global region-of-interest threshold value tau =0,1,2.
SetEnTime={EnTime(τ(1)),EnTime(τ(2)),...,EnTime(τ(255))} (2)
Step three, sequentially calculating a peak signal-to-noise ratio between the region-of-interest block ciphertext image of the medical image Pimg and the original image by using a global region-of-interest threshold value tau =0,1,2.
SetEnPSNR={EnPSNR(τ(1)),EnPSNR(τ(2)),...,EnPSNR(τ(255))} (3)
EnPSNR(τ)=psnr(EnImg(τ),Pimg)
Wherein
Wherein psnr is a function representing a peak signal-to-noise ratio, MSE represents a mean square error of the current image I and the reference image I2, and H, W are the height and the width of the images respectively; l is the number of bits per pixel, and is typically 8, i.e., the number of pixel gray levels is 256. The distribution of the peak SNR set as a function of the threshold is shown in FIG. 2B.
Fourthly, sequentially calculating the structural similarity between the region-of-interest block ciphertext image of the medical image Pimg and the original image by using a global region-of-interest threshold value tau =0,1,2,.. 254 to obtain a structural similarity set:
SetEnSSIM={EnSSIM(τ(1)),EnSSIM(τ(2)),...,EnSSIM(τ(255))} (4)
EnSSIM(τ)=ssim(EnImg(τ),Pimg)
ssim(X1,X2)=l(X1,X2)c(X1,X2)s(X1,X2)
where ssim is a function of structural similarity, μ X1 、μ X2 Denotes the mean, σ, of images X1 and X2, respectively X1 、σ X2 Denotes the variance, σ, of the images X1 and X2, respectively X1X2 Represents the covariance, C, of images X1 and X2 1 、C 2 、C 3 As a constant, the distribution of the structural similarity set over the threshold variation profile is shown in fig. 2C.
Step four, sequentially calculating the information entropy of the region-of-interest block ciphertext image of the medical image Pimg by using a global region-of-interest threshold value tau =0,1,2.
SetEnEntropy={EnEntropy(τ(1)),EnEntropy(τ(2)),...,EnEntropy(τ(255))) (5)
EnEntropy(τ)=infoEntropy(EnImg(τ))
Where infoEntrol is the information entropy function, p (φ) i ) Indicates a random event phi of phi i L represents the pixel depth of the image, and the distribution graph of the entropy set of the ciphertext information along with the change of the threshold is shown in fig. 2D.
Step four, calculating to obtain an encryption effect set according to the encryption performance sets in the step four, the step two, the step four and the step five:
SetEnEffect={EnEffect(τ(1),EnEffect(τ(2)),...,EnEEffect(τ(255))
EnEffect(τ)=1/log(ω 1 P 1 EnTime(τ)+ω 2 EnPSNR(τ)+ω 3 P 2 EnSSIM(τ)+ω 4 EnEntropy(τ)) (6)
wherein: omega 1 ,ω 2 ,ω 3 ,ω 4 For each encryption performance set weight, omega, from step four two to step four five 1 As a weight of the encryption time, omega 2 As peak signal-to-noise ratio weight, ω 3 As a structural similarity weight, ω 4 Is the information entropy weight; p 1 ,P 2 To enlarge the factor, P 1 For encrypting the time amplification factor, P 2 Is the magnification factor of the structural similarity. In this example ω 1 =0.6,ω 2 =0.2,ω 3 =ω 4 =0.1,P 1 =20,P 2 =10. The set of cryptographic effects as a function of the threshold is shown in fig. 2E.
Step five, fitting the encryption effect set SetEnEffect by using a machine-learned polynomial linear regression method to obtain an encryption effect gain function SetEnEffect Income, wherein the specific fitting method is as follows:
fifthly, amplifying the encryption effect set SetEnEffect in the fourth step in proportion to obtain SetEnEffect' which is used as a result set of polynomial linear regression.
Step five two, taking X =1,2.
And fifthly, increasing the high-order item of X to obtain a high-order item input set Poly _ X.
poly_reg=PolynomialFeatures(degree=3) (8)
Poly_X=poly_reg.fit_transformn(X)
Where PolynomialFeatures denotes the term function that sets the input X, and poly _ reg.fit _ transform denotes the conversion of a one-dimensional feature to a multi-dimensional feature, in this example a 3 rd order polynomial.
And fifthly, approximating the result set SetEnEffect' by adopting a polynomial regression method.
LinearRegression().fit(Poly_X,SetEnEffect) (9)
Wherein Linear regression represents a linear regression method, and the obtained results are shown in FIG. 3.
And fifthly, obtaining each item coefficient of the polynomial according to the fitting result to obtain an encryption effect gain function SetEneffectIncome. In this example, setEneffectosome =8.14765957e-02 x-1.83506128e-03 x ^2+2.64347524e-06 x, 3+ 168.18614381248.
Step six, when the participant ROI i If no strategy is executed, an encryption scheme of the global image is adopted, the encryption performance from the fourth step to the fourth step is calculated, and the initial profit value SetEneffiecomeme of the encryption effect profit function SetEneffiecomeme is obtained 0 . Wherein min (SetEneffect') > SetEneffect Income 0 Min () is the function that finds the minimum element value of the set. In this example, the initial profit value SetEnEffectIncome 0 Is 60.25.
Step seven, the game Nash bargaining price counter-offer solution of the area of interest is expressed as:
obtaining optimal solution tau of threshold game of interest region by using Lagrange multiplier method optinal The optimal solution in this example is 23.
And step eight, carrying out region-of-interest division on the original image Pimg by using the region-of-interest threshold game optimal solution 23 in the step seven, and encrypting the region-of-interest to obtain an encrypted image. The ciphertext image may be as shown in fig. 4B.
Claims (2)
1. The machine learning-based interesting region parameter game analysis method is realized by the following steps:
taking an L-bit gray level medical image with the size of M multiplied by N as an original medical image Pimg, wherein L is the pixel depth of the image;
step two, setting the number of game participants as P and the participants as ROI i I =1,2,3, P selects different region of interest partitioning parameters to partition the region of interest, where the participant ROI i The specific method for dividing the region of interest is as follows:
step two, dividing the original medical image Pimg into image blocks with the size of N multiplied by N to obtain M multiplied by N/(N multiplied by N) image blocks
Step two, sequentially calculating the image blocks in the step twoAverage of the gray values of all pixels within->
Step two and three, the participant ROI i The discrimination method of the region of interest is as follows:
when in useWhen it is correct, the image block is>Setting a corresponding region-of-interest zone bit for the region of interestIs 1;
when in useIf so, then the image block is->Setting a corresponding region of interest flag for a region of non-interest>Is 0;
where t _ ROI i ROI for participants i Threshold of region of interest, t _ ROI i ∈[0,2 L -1) region of interest flagBlock 1 indicates participant ROI i The region of interest block, namely the selected encryption region block;
fourthly, performing encryption performance analysis on the region of interest according to a global region of interest threshold tau to obtain an encryption effect set, wherein the global region of interest threshold tau belongs to the 0,2 L -1), the specific analysis steps are:
step four, sequentially encrypting the region-of-interest blocks of the original medical image Pimg by using the global region-of-interest threshold tau to obtain a ciphertext image set SetEnImg;
step two, recording the encryption time of the region-of-interest block for encrypting the original medical image Pimg by using the global region-of-interest threshold tau in sequence to obtain an encrypted image time set SetEnTime;
step four, sequentially calculating a peak signal-to-noise ratio between the region-of-interest block ciphertext image of the original medical image Pimg and the original medical image encrypted by the global region-of-interest threshold tau to obtain a peak signal-to-noise ratio set SetEnPSNR;
fourthly, sequentially calculating the structural similarity between the region-of-interest block ciphertext image of the original medical image Pimg and the original medical image encrypted by the global region-of-interest threshold tau to obtain a structural similarity set SetEnSSIM;
step four, sequentially calculating the information entropy of the region-of-interest block ciphertext image obtained by encrypting the original medical image Pimg by using the global region-of-interest threshold value to obtain a ciphertext information entropy set SetEntrack;
step four, calculating and obtaining an encryption effect set SetEnEffect according to the encryption image time set SetEnTime, the peak signal-to-noise ratio set SetEnPSNR, the structural similarity set SetEnSSIM and the ciphertext information entropy set SetEncopy described in the step four to the step four; represented by the formula:
SetEnEffect={EnEffect(τ(1)),EnEffect(τ(2)),...,EnEEffect(τ(2 L -1))}
EmEffect(τ)=1/log(ω 1 P 1 EnTime(τ)+ω 2 EnPSNR(τ)+ω 3 P 2 EnSSIM(τ)+ω 4 EnEntropy(τ)) (6)
wherein, ω is 1 As a weight of the encryption time, omega 2 As peak signal-to-noise ratio weight, ω 3 Is a structural phaseSimilarity weight, ω 4 Is an information entropy weight; p 1 For encrypting the time amplification factor, P 2 The amplification factor is the structural similarity;
step five, fitting the encryption effect set SetEnEffect by adopting a machine-learning polynomial linear regression method to obtain an encryption effect gain function SetEnEffect Income;
step six, when the participant ROI i If no strategy is executed, the encryption performance of the step four to the step four is calculated, and the initial profit value SetEneffiecomes of the encryption effect profit function SetEneffiecomes is obtained 0 ;
Setting min (SetEneffett') > SetEneffetcIncome 0 Min () is the function of solving the minimum element value of the set;
step seven, obtaining the initial profit value SetEneffectIncome of the encryption effect profit function SetEneffectIncome according to the step six 0 Obtaining a game Nash bargaining counter-price solution of the area of interest as follows:
obtaining optimal solution tau of the area of interest threshold game by adopting Lagrange multiplier method optimal ;
Step eight, adopting the area-of-interest threshold game optimal solution tau in the step seven optimal And dividing the region of interest of the original medical image Pimg, and encrypting the region of interest to obtain an encrypted image.
2. The machine learning based region of interest parametric game analysis method of claim 1, wherein: in the fifth step, the concrete fitting method is as follows:
fifthly, amplifying the encryption effect set SetEnEffect in the fourth step in proportion to obtain an amplification result SetEnEffet, and taking the amplification result SetEnEffet as a result set of polynomial linear regression;
step five two, X =1,2 L -1 as an input set;
fifthly, increasing the high-order item of X to obtain a high-order item input set Poly _ X;
fifthly, fitting the amplified result set SetEnEffect by adopting a polynomial regression method to obtain a fitting result;
and fifthly, obtaining each item coefficient of the polynomial according to the fitting result, and obtaining an encryption effect gain function SetEneffectIncome.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201911356216.0A CN111048185B (en) | 2019-12-25 | 2019-12-25 | Interesting region parameter game analysis method based on machine learning |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201911356216.0A CN111048185B (en) | 2019-12-25 | 2019-12-25 | Interesting region parameter game analysis method based on machine learning |
Publications (2)
Publication Number | Publication Date |
---|---|
CN111048185A CN111048185A (en) | 2020-04-21 |
CN111048185B true CN111048185B (en) | 2023-03-28 |
Family
ID=70240269
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201911356216.0A Active CN111048185B (en) | 2019-12-25 | 2019-12-25 | Interesting region parameter game analysis method based on machine learning |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN111048185B (en) |
Families Citing this family (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113051357B (en) * | 2021-03-08 | 2022-09-30 | 中国地质大学(武汉) | Vector map optimization local desensitization method based on game theory |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
KR101444828B1 (en) * | 2014-04-30 | 2014-09-26 | 동국대학교 산학협력단 | Method for storing encrypted image and searching the image |
EP3207480A2 (en) * | 2014-10-15 | 2017-08-23 | Trice Imaging, Inc. | Systems and methods for encrypting, converting and interacting with medical images |
CN108305671A (en) * | 2018-01-23 | 2018-07-20 | 深圳科亚医疗科技有限公司 | By computer implemented medical image dispatching method, scheduling system and storage medium |
CN108650434A (en) * | 2018-05-08 | 2018-10-12 | 吉林大学 | A kind of method of image encryption |
CN108665964A (en) * | 2018-05-14 | 2018-10-16 | 江西理工大学应用科学学院 | A kind of medical image wavelet field real-time encryption and decryption algorithm based on multi-chaos system |
CN109492416A (en) * | 2019-01-07 | 2019-03-19 | 南京信息工程大学 | A kind of guard method of big data image and system based on safety zone |
Family Cites Families (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2006070249A1 (en) * | 2004-12-27 | 2006-07-06 | Emitall Surveillance S.A. | Efficient scrambling of regions of interest in an image or video to preserve privacy |
-
2019
- 2019-12-25 CN CN201911356216.0A patent/CN111048185B/en active Active
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
KR101444828B1 (en) * | 2014-04-30 | 2014-09-26 | 동국대학교 산학협력단 | Method for storing encrypted image and searching the image |
EP3207480A2 (en) * | 2014-10-15 | 2017-08-23 | Trice Imaging, Inc. | Systems and methods for encrypting, converting and interacting with medical images |
CN108305671A (en) * | 2018-01-23 | 2018-07-20 | 深圳科亚医疗科技有限公司 | By computer implemented medical image dispatching method, scheduling system and storage medium |
CN108650434A (en) * | 2018-05-08 | 2018-10-12 | 吉林大学 | A kind of method of image encryption |
CN108665964A (en) * | 2018-05-14 | 2018-10-16 | 江西理工大学应用科学学院 | A kind of medical image wavelet field real-time encryption and decryption algorithm based on multi-chaos system |
CN109492416A (en) * | 2019-01-07 | 2019-03-19 | 南京信息工程大学 | A kind of guard method of big data image and system based on safety zone |
Non-Patent Citations (5)
Title |
---|
A game theory-based block image compression method in encryption domain;S. Liu et al.;《The Journal of Supercomputing》;第71卷;第3353-3372页 * |
Watermarking and Encryption in Medical Image Through Roi-Lossless Compression;S. Nithya et al.,;《International Conference on Communication and Signal Processing》;第1-5页 * |
基于多混沌和分数Fourier的光学图像加密算法;底晓强等;《南京大学学报(自然科学)》;第55卷(第2期);第251-263页 * |
基于提升小波的Rol图像信息隐藏研究与实现;张力;《中国优秀硕士学位论文全文数据库(电子期刊) 信息科技辑》(第5期);第I138-97页 * |
数字图像隐写容量的博弈分析;蒋翠玲等;《微电子学与计算机》;第27卷(第1期);第22-24+28页 * |
Also Published As
Publication number | Publication date |
---|---|
CN111048185A (en) | 2020-04-21 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Balasamy et al. | A fuzzy based ROI selection for encryption and watermarking in medical image using DWT and SVD | |
Liu et al. | Privacy-preserving object detection for medical images with faster R-CNN | |
Balasamy et al. | Feature extraction-based medical image watermarking using fuzzy-based median filter | |
Memon et al. | Hybrid watermarking of medical images for ROI authentication and recovery | |
Ou et al. | High capacity reversible data hiding based on multiple histograms modification | |
Srikanth et al. | Multilevel thresholding image segmentation based on energy curve with harmony Search Algorithm | |
Shiri et al. | Decentralized distributed multi-institutional PET image segmentation using a federated deep learning framework | |
CN115378574B (en) | Lightweight dynamic image data encryption method and system | |
Balasamy et al. | Improving the security of medical image through neuro-fuzzy based ROI selection for reliable transmission | |
Yang et al. | A clustering-based framework for improving the performance of JPEG quantization step estimation | |
Zhang et al. | An efficient multi-level encryption scheme for stereoscopic medical images based on coupled chaotic system and Otsu threshold segmentation | |
CN112508782A (en) | Network model training method, face image super-resolution reconstruction method and equipment | |
CN111988144B (en) | DNA one-time pad image encryption method based on multiple keys | |
Ahmad et al. | Hiding patients’ medical reports using an enhanced wavelet steganography algorithm in DICOM images | |
Al-Dmour et al. | An efficient steganography method for hiding patient confidential information | |
CN113160944A (en) | Medical image sharing method based on block chain | |
CN111048185B (en) | Interesting region parameter game analysis method based on machine learning | |
Liao et al. | DeepWSD: Projecting degradations in perceptual space to wasserstein distance in deep feature space | |
Gong et al. | Robust and secure zero-watermarking algorithm for medical images based on Harris-SURF-DCT and chaotic map | |
Bu et al. | 3D conditional generative adversarial network‐based synthetic medical image augmentation for lung nodule detection | |
Yang et al. | Color image steganalysis based on embedding change probabilities in differential channels | |
Singh et al. | Blind and secured adaptive digital image watermarking approach for high imperceptibility and robustness | |
Nguyen et al. | A new chapter for medical image generation: The stable diffusion method | |
Xing et al. | Brain MR atlas construction using symmetric deep neural inpainting | |
Singh et al. | Intelligent Data Security Solutions for e-Health Applications |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |