CN109544656A - A kind of compressed sensing image rebuilding method and system based on generation confrontation network - Google Patents
A kind of compressed sensing image rebuilding method and system based on generation confrontation network Download PDFInfo
- Publication number
- CN109544656A CN109544656A CN201811404831.XA CN201811404831A CN109544656A CN 109544656 A CN109544656 A CN 109544656A CN 201811404831 A CN201811404831 A CN 201811404831A CN 109544656 A CN109544656 A CN 109544656A
- Authority
- CN
- China
- Prior art keywords
- generator
- discriminator
- original image
- network model
- 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.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 22
- 238000005259 measurement Methods 0.000 claims abstract description 51
- 238000012549 training Methods 0.000 claims abstract description 36
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 33
- 238000005070 sampling Methods 0.000 claims abstract description 17
- 238000013528 artificial neural network Methods 0.000 claims abstract description 6
- 230000006870 function Effects 0.000 claims description 40
- 239000011159 matrix material Substances 0.000 claims description 18
- 230000004913 activation Effects 0.000 claims description 11
- 235000004257 Cordia myxa Nutrition 0.000 claims description 4
- 244000157795 Cordia myxa Species 0.000 claims description 4
- 238000010276 construction Methods 0.000 claims description 4
- 238000005457 optimization Methods 0.000 claims description 3
- 238000013507 mapping Methods 0.000 abstract description 2
- 230000009977 dual effect Effects 0.000 abstract 1
- 230000000694 effects Effects 0.000 description 5
- 238000012360 testing method Methods 0.000 description 5
- 238000010586 diagram Methods 0.000 description 4
- 238000003384 imaging method Methods 0.000 description 2
- 238000004445 quantitative analysis Methods 0.000 description 2
- 238000004364 calculation method Methods 0.000 description 1
- 238000007906 compression Methods 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 230000010365 information processing Effects 0.000 description 1
- 230000010354 integration Effects 0.000 description 1
- 238000012804 iterative process Methods 0.000 description 1
- ONCZDRURRATYFI-QTCHDTBASA-N methyl (2z)-2-methoxyimino-2-[2-[[(e)-1-[3-(trifluoromethyl)phenyl]ethylideneamino]oxymethyl]phenyl]acetate Chemical compound CO\N=C(/C(=O)OC)C1=CC=CC=C1CO\N=C(/C)C1=CC=CC(C(F)(F)F)=C1 ONCZDRURRATYFI-QTCHDTBASA-N 0.000 description 1
- 238000004451 qualitative analysis Methods 0.000 description 1
- 238000004088 simulation Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T11/00—2D [Two Dimensional] image generation
-
- 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
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
- Y02T10/00—Road transport of goods or passengers
- Y02T10/10—Internal combustion engine [ICE] based vehicles
- Y02T10/40—Engine management systems
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Biophysics (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Health & Medical Sciences (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a kind of based on the compressed sensing image rebuilding method for generating confrontation network, it include: that S1, the measurement vector sampled according to original image and reconstruction image size construct generation confrontation network model neural network based, and the objective function designed for optimizing the generation confrontation network model parameter;S2, the default training parameter generated when fighting network model;S3, according to the objective function, using back-propagation algorithm alternately training generator and discriminator;If S4, generation confrontation network model convergence, trained network can be directly realized by compressed sensing task, and model output is by the correspondence original image for measuring vector reconstruction and going out;Otherwise S2-S4 is returned to step.The present invention utilizes the powerful mapping ability of generator, and preliminary reconstruction original image has achieved the purpose that Exact Reconstruction original image under low sampling rate so that the image pixel that generator is rebuild is distributed closer to original image using the dual training of generator and discriminator.
Description
Technical Field
The invention relates to the technical field of image information processing, in particular to a compressive sensing image reconstruction method and a compressive sensing image reconstruction system based on a generation countermeasure network.
Background
Compressed Sensing (CS) is a novel signal acquisition theory, combines traditional sampling and compression processes, can directly acquire measurement data far below the Nyquist sampling rate, can reduce sampling cost and storage resources, and meanwhile, the encoding end of the Compressed Sensing model only needs to perform linear random measurement, and the complex optimization process of reconstructing signals is completed at the decoding end.
The imaging systems are based on a compressed sensing theory, and an iterative optimization algorithm is used for reconstructing images for a small number of observed measured values. However, these reconstruction algorithms require complex iterative operations, the reconstruction time is long, and the quality of the reconstructed image is poor at a low sampling rate, which hinders the deep development and industrial application of compressed sensing imaging.
Disclosure of Invention
The purpose of the invention is as follows: in order to overcome the defects of the prior art, the invention provides a compressed sensing image reconstruction method based on a generation countermeasure network, which can solve the problems of long reconstruction time and poor quality of reconstructed images, and also provides a compressed sensing image reconstruction system based on the generation countermeasure network.
The technical scheme is as follows: the invention discloses a compressed sensing image reconstruction method based on a generation countermeasure network, which comprises the following steps:
s1, constructing a generated confrontation network model based on a neural network according to the measurement vector obtained by image sampling and the size of a reconstructed image, and designing an objective function for optimizing parameters of the generated confrontation network model;
s2, presetting parameters for training the generation of the confrontation network model;
s3, alternately training the generator and the discriminator by adopting a back propagation algorithm according to the target function;
s4, if the generated confrontation network model is converged, the trained network can directly realize a compressed sensing task, and the model is output as a corresponding original image reconstructed by the measurement vector; otherwise, the execution returns to the steps S2-S4.
Preferably, in S1, the measurement vector is expressed as:
y=Φx+ξ,y∈RM,Φ∈RM×N,x∈RN
wherein y represents a measurement vector, Φ represents a measurement matrix, x represents a data row vector which is formed after the data matrix of the image to be sampled is changed into vectorization, M represents the size of the measurement vector, and N represents the number of pixels of the image to be sampled.
Preferably, in S1, the formula for generating the confrontation network model is as follows:
J(D,G)=minGmaxD[lrec(G(y))+lreg(G(y))+lGAN(G(y),D)]
g is a generator, D is a discriminator, y is a measurement vector, x represents a data row vector formed by vectorizing the data matrix of the image to be sampled, and lrecTo reconstruct the loss,/regFor total variation regularization, lGANIs the countermeasure loss of the discriminator network.
Preferably, the reconstruction loss, the total variation regularization and the corresponding formula of the countermeasure loss of the discriminator network are substituted into the formula for generating the countermeasure network model, so as to respectively obtain the objective functions for optimizing the parameters of the generator and the discriminator;
the objective function of the generator is represented as:
the objective function of the discriminator is:
wherein,for the reconstructed image of the generator, i and j are the pixel positions of the original image, pdataAnd pyIs the statistical distribution of the original image and the random measurement vector, and gamma is the loss weight of the discriminator.
Preferably, in S3, the activation function used when training the generator is Selu, and the activation function used when training the discriminator is Lrelu.
The invention also discloses a system for realizing the compressed sensing image reconstruction method based on the generation countermeasure network, which comprises the following steps:
the network model construction module is used for constructing a generation countermeasure network model based on a neural network according to the measurement vector obtained by image sampling and the size of a reconstructed image, and designing an objective function for optimizing parameters of the generation countermeasure network model;
the parameter presetting module is used for presetting parameters when the confrontation network model is generated;
the generator training module is used for alternately training the generator and the discriminator by adopting a back propagation algorithm according to the target function;
the convergence judging module is used for judging whether the generated confrontation network model converges or not, if the generated confrontation network model converges, the trained network can directly realize a compressed sensing task, and the model output is a corresponding original image reconstructed by the measurement vector; otherwise, returning to the parameter presetting module.
Preferably, in the network model building module,
the measurement vector is represented as:
y=Φx+ξ,y∈RM,Φ∈RM×N,x∈RN
wherein y represents a measurement vector, Φ represents a measurement matrix, x represents a data row vector which is formed after the data matrix of the image to be sampled is changed into vectorization, M represents the size of the measurement vector, and N represents the number of pixels of the image to be sampled.
Preferably, in the network model building module, the formula for generating the confrontation network model is as follows:
J(D,G)=minGmaxD[lrec(G(y))+lreg(G(y))+lGAN(G(y),D)]
wherein G is a generator, D is a discriminator, y is a measurement vector, x represents a data row vector which is formed by converting a data matrix of the original image into vectorization, and l is a data row vector which is formed by converting the data matrix of the original image into vectorizationrecTo reconstruct the loss,/regFor total variation regularization, lGANIs the countermeasure loss of the discriminator network.
Preferably, the reconstruction loss, the total variation regularization and the corresponding formula of the countermeasure loss of the discriminator network are substituted into the formula for generating the countermeasure network model, so as to respectively obtain the objective functions for optimizing the parameters of the generator and the discriminator;
the objective function of the generator is represented as:
the objective function of the discriminator is:
wherein,for the reconstructed image of the generator, i and j are the pixel positions of the original image, pdataAnd pyIs the statistical distribution of the original image and the random measurement vector, and gamma is the loss weight of the discriminator.
Preferably, in the generator training module, the activation function used when training the generator is Selu, and the activation function used when training the discriminator is Lrelu.
Has the advantages that: the invention utilizes a compressed sensing theory to establish a compressed sensing image reconstruction method based on a generation countermeasure network, a model initially reconstructs an original image by utilizing the strong mapping capability of a generator under the condition of low image sampling rate, and simultaneously, the capability of the generator is further enhanced by utilizing the countermeasure training of the generator and a discriminator, so that the pixel distribution of the image reconstructed by the generator is closer to the original image, and the aim of accurately reconstructing the original image under the condition of low sampling rate is fulfilled.
Drawings
FIG. 1 is a flow chart of a method of the present invention;
fig. 2 is a comparison result between a reconstructed image of a CelebA face data set and an original image by a plurality of algorithms according to the present invention, wherein fig. 2a is a schematic diagram of a comparison result between a reconstructed image and an original image obtained by a ReconNet algorithm, fig. 2b is a comparison result between a reconstructed image and an original image obtained by a CSGAN algorithm, fig. 2c is a schematic diagram of a comparison result between a reconstructed image and an original image obtained by a DCGAN algorithm, and fig. 2d is a diagram of a comparison result between a reconstructed image and an original image obtained by a method according to the present invention;
FIG. 3 is a graph of the comparison result of the mean square error between the reconstructed image and the original image according to the Reconnet algorithm, the CSGAN algorithm and the DCGAN algorithm of the present invention;
FIG. 4 is a graph of the comparison results between the Reconfigut algorithm, the CSGAN algorithm, the DCGAN algorithm and the reconstructed image obtained by the present algorithm and the original image in the peak SNR and the image structure similarity index;
FIG. 5 is a graph of the Reconnet algorithm, CSGAN algorithm, DCGAN algorithm and the time required for the algorithm to reconstruct an image in accordance with the present invention;
fig. 6 is a schematic diagram of the system structure according to the present invention.
Detailed Description
As shown in fig. 1, the present invention provides a compressed sensing image reconstruction method based on a generative countermeasure network, including:
step 1, sampling an original image to obtain a measurement vector
The measurement vector y ═ Φ x + ξ, y ∈ RM,Φ∈RM×N,x∈RNY represents a measurement vector, phi represents a measurement vector, x represents a data row vector formed by vectorizing the picture data matrix, M represents the size of the measurement vector, and N represents the number of pixels of the original image.
And 2, constructing and generating a confrontation network model according to the size of the measurement vector and the size of the reconstructed image, and designing an objective function for optimizing parameters of the network model.
The size of the vector is measured to determine the size of the input of the generator, and the size of the reconstructed image determines the final output size of the generator. The generation of the confrontation network model is as follows:
J(D,G)=minGmaxD[lrec(G(y))+lreg(G(y))+lGAN(G(y),D)]
wherein G is the generator, D is the discriminator, and the reconstruction loss isTotal variation regularizationThe countermeasure loss of the discriminator network isSubstituting the reconstruction loss, the total variation regularization and the confrontation loss of the discriminator network into the network model to obtain objective functions respectively optimizing the parameters of the generator and the discriminator,respectively, as an objective function of the generatorThe objective function of the discriminator is For the reconstructed image of the generator, i and j are the pixel positions of the original image, pdataAnd pyIs the statistical distribution of the original image and the random measurement vector, and gamma is the loss weight of the discriminator.
Step 3, presetting network model training hyperparameters
The preset parameters include model learning rate α during training, iteration number L, training batch size S, loss weight γ of the discriminator during generator training, depth and number of layers of the generator and discriminator network, activation function of the generator and generator category of the discriminator.
And 4, alternately training the generator and the discriminator by adopting a back propagation algorithm according to the target function.
Specifically, step 4 specifically includes:
step 41, selecting S training images { x in training set(1),…,x(s)Get the corresponding measurement vector y(1),…,y(s)};
And 42, updating the parameter omega in the discriminator through a back propagation algorithm.
First, discriminator loss is calculated
Second, fix the generator parameter θ, update the discriminator parameter ω ← ω + α · RMSProp (ω, d)ω). Wherein RMSProp is gradient descent algorithmOne kind of (1).
Step 43, constructing a generator loss function required for training the generator network according to the modelCalculating loss function of S training images in batch, fixing discriminator parameter omega, and updating generator parameter theta by RMSProp gradient descent algorithm, theta ← theta- α, RMSProp (theta, g)θ)。
And step 44, sequentially performing step 41, step 42 and step 43 on all images of the whole training set, and performing L iterations in total.
And 5, if the network model is converged, the trained network can directly realize the compressed sensing task of the segment opposite end, the output of the model is the corresponding original image reconstructed by the measurement vector, otherwise, the step 3 is executed, and the step 3 to the step 5 are circulated.
Specifically, the step 5 specifically includes:
and step 51, judging whether the generator network and the discriminator network are converged. During the iterative process of network training, when the discriminator loses dωSum generator gθWhen the values are all reduced and respectively asymptotic to a certain value, the network is judged to be converged.
Step 52, after the measurement vector obtained by sampling the measurement matrix Φ is input into the converged generator network, the output of the generator is the corresponding reconstructed image.
And step 53, if the iterative training is not converged, returning to execute the step 3.
In order to verify the effect of the present invention, a simulation experiment is performed on the present invention, the test image specification is 64 × 64, a model is trained and tested on the CelebA face data set, relevant parameters are set, the model learning rate α is 0.002, the iteration number L is 30, the number S of pictures input each time is 16, γ is 0.01, and the process from input to output of the network of the generator and the discriminator is shown in table 1 and table 2 below.
Table 1 training depth and activation function of the Generator
TABLE 2 training depth and activation function of discriminator
The evaluation of the experiment used both qualitative and quantitative analytical methods.
Fig. 2 shows the image reconstruction effect comparison of the present invention and the ReconNet, CSGAN, and DCGAN algorithms at sample block numbers of 20, 100, and 500, respectively. The sampling block is the size of the measurement vector y, and as can be seen from fig. 2, the reconstruction effect of the invention is obviously superior to the three algorithms of ReconNet, CSGAN and DCGAN for the same four face images.
For quantitative analysis comparison, MSE, PSNR and SSIM are used to evaluate image quality, and reconstruction time (time unit is millisecond) of a single image is used to evaluate reconstruction speed of the algorithm. Where MSE is Mean square error (Mean squared error), i.e. average error of a single pixel in an image, PSNR is Peak Signal to Noise Ratio (Peak Signal to Noise Ratio), SSIM is structural similarity index (structural similarity index), and the calculation is as follows:
wherein range represents the dynamic range of image pixel values, μ is the mean, σ is the variance, c1=(k1L)2,c1=(k1L)2Is a constant for maintaining stability, L is the dynamic range of pixel values, k1=0.01,k2=0.03。
When quantitative comparison is carried out, 4 test pictures are selected from a test set, sampled by using a measurement matrix phi respectively and input into each model, a reconstructed image is calculated and output through the model, the reconstructed image is compared with the corresponding original image, and corresponding MSE (mean square error), PSNR (signal to noise ratio) and SSIM (small scale integration) values are calculated.
FIG. 3 shows the MSE (under log2 scaling) comparison values of the algorithm of the present invention and several algorithms Reconnet, CSGAN, DCGAN, TWLST, LASSO in the CelebA test set, respectively, with Number of Measurements (Number of samples) on the abscissa and the error per pixel, i.e., the mean square error (Reconstruction error per pixel) on the ordinate; fig. 4 shows the PSNR and SSIM comparison values for different sample blocks for the algorithm of the present invention and ReconNet, CSGAN, DCGAN. FIG. 5 shows a comparison of the average time (in milliseconds) for reconstructing an image in the mnist, fmnist and CelebA test sets for the present algorithm and several of Reconnet, CSGAN, DCGAN, TWIST and LASSO, respectively.
The invention also provides a compressed sensing image reconstruction system based on the generation countermeasure network, as shown in fig. 6, including:
the network model construction module is used for constructing a generation countermeasure network model based on a neural network according to the measurement vector obtained by image sampling and the size of a reconstructed image, and designing an objective function for optimizing parameters of the generation countermeasure network model;
the parameter presetting module is used for presetting parameters when the confrontation network model is generated;
the generator training module is used for alternately training the generator and the discriminator by adopting a back propagation algorithm according to the target function;
the convergence judging module is used for judging whether the generated confrontation network model converges or not, if the generated confrontation network model converges, the trained network can directly realize a compressed sensing task, and the model output is a corresponding original image reconstructed by the measurement vector; otherwise, returning to the parameter presetting module.
The system is realized based on a compressed sensing image reconstruction method for generating a countermeasure network, the specific technology is similar to the method, and the invention is not repeated herein. Although the preferred embodiments of the present invention have been described, the scope of the present invention is not limited thereto, and any changes or substitutions that can be easily made by those skilled in the art within the technical scope of the present invention should be covered by the scope of the present invention
In conclusion, compared with the conventional compressed sensing, the reconstruction method has the problems of long iteration time and poor reconstruction quality under a low sampling rate. The invention utilizes deep learning to establish a compressed sensing image reconstruction model based on a generation countermeasure network, enables a generator G to reconstruct an image, improves the pixel distribution of the reconstructed image by using a discriminator D, reduces reconstruction errors, has better reconstruction effect on complex image reconstruction under low sampling rate, has better reconstruction effect on more complex color images, and has certain advantages in reconstruction error, similarity and efficiency or visual effect.
Claims (10)
1. A compressed sensing image reconstruction method based on a generation countermeasure network is characterized by comprising the following steps:
s1, constructing a generated confrontation network model based on a neural network according to the measurement vector obtained by sampling the original image and the size of the reconstructed image, and designing an objective function for optimizing parameters of the generated confrontation network model;
s2, presetting parameters for training the generation of the confrontation network model;
s3, alternately training the generator and the discriminator by adopting a back propagation algorithm according to the target function;
s4, if the generated confrontation network model is converged, the trained network can directly realize a compressed sensing task, and the model is output as a corresponding original image reconstructed by the measurement vector; otherwise, the execution returns to the steps S2-S4.
2. The method for reconstructing a compressed sensing image based on a generative countermeasure network according to claim 1, wherein in the step S1, the measurement vector is represented as:
y=Φx+ξ,y∈RM,Φ∈RM×N,x∈RN
wherein y represents a measurement vector, Φ is a measurement matrix, x represents a data row vector which is formed by converting the data matrix of the original image into vectorization, M represents the size of the measurement vector, and N is the number of pixels of the original image.
3. The method for reconstructing a compressed sensing image based on a generative confrontation network as claimed in claim 1, wherein in S1, the formula for generating the confrontation network model is:
J(D,G)=minGmaxD[lrec(G(y))+lreg(G(y))+lGAN(G(y),D)]
wherein G is a generator, D is a discriminator, y is a measurement vector, x represents a data row vector which is formed by converting a data matrix of the original image into vectorization, and l is a data row vector which is formed by converting the data matrix of the original image into vectorizationrecTo reconstruct the loss,/regFor total variation regularization, lGANIs the countermeasure loss of the discriminator network.
4. The compressive sensing image reconstruction method based on the generative confrontation network as claimed in claim 3, wherein the reconstruction loss, total variation regularization and confrontation loss corresponding formula of the discriminator network are substituted into the formula of the generative confrontation network model to respectively obtain an objective function of optimizing parameters of a generator and a discriminator;
the objective function of the generator is represented as:
the objective function of the discriminator is:
wherein,for the reconstructed image of the generator, i and j are the pixel positions of the original image, pdataAnd pyIs the statistical distribution of the original image and the random measurement vector, and gamma is the loss weight of the discriminator.
5. The method for reconstructing a compressed sensing image based on a generative countermeasure network according to claim 1, wherein in the step S3, an activation function used in training the generator is Selu, and an activation function used in training the discriminator is Lrelu.
6. A system implemented according to the method for reconstructing a compressed sensing image based on a generative countermeasure network according to claims 1 to 5, comprising:
the network model construction module is used for constructing a generated confrontation network model based on a neural network according to a measurement vector obtained by sampling an original image and the size of a reconstructed image, and designing an objective function for optimizing parameters of the generated confrontation network model;
the parameter presetting module is used for presetting parameters when the confrontation network model is generated;
the generator training module is used for alternately training the generator and the discriminator by adopting a back propagation algorithm according to the target function;
the convergence judging module is used for judging whether the generated confrontation network model converges or not, if the generated confrontation network model converges, the trained network can directly realize a compressed sensing task, and the model output is a corresponding original image reconstructed by the measurement vector; otherwise, returning to the parameter presetting module.
7. The generation-based countermeasure network-oriented compressed sensing image reconstruction system of claim 6, wherein the network model building module,
the measurement vector is represented as:
y=Φx+ξ,y∈RM,Φ∈RM×N,x∈RN
wherein y represents a measurement vector, Φ represents a measurement matrix, x represents a data row vector which is formed after the data matrix of the original image is changed into vectorization, M represents the size of the measurement vector, and N represents the number of pixels of the image to be sampled.
8. The system for reconstructing a compressed sensing image based on generation of a countermeasure network according to claim 6, wherein the formula for generating the countermeasure network model in the network model construction module is as follows:
J(D,G)=minGmaxD[lrec(G(y))+lreg(G(y))+lGAN(G(y),D)]
wherein G is a generator, D is a discriminator, y is a measurement vector, x represents a data row vector which is formed by converting a data matrix of the original image into vectorization, and l is a data row vector which is formed by converting the data matrix of the original image into vectorizationrecTo reconstruct the loss,/regFor total variation regularization, lGANIs the countermeasure loss of the discriminator network.
9. The compressive sensing image reconstruction system based on generation countermeasure network of claim 8, characterized in that the reconstruction loss, total variation regularization and countermeasure loss corresponding formulas of discriminator network are substituted into the formula of generation countermeasure network model to obtain objective functions of optimization generator and discriminator parameters, respectively;
the objective function of the generator is represented as:
the objective function of the discriminator is:
wherein,for the reconstructed image of the generator, i and j are the pixel positions of the original image, pdataAnd pyIs the statistical distribution of the original image and the random measurement vector, and gamma is the loss weight of the discriminator.
10. The system according to claim 6, wherein the generator training module trains the generator with Selu as an activation function and Lrelu as an activation function.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811404831.XA CN109544656B (en) | 2018-11-23 | 2018-11-23 | Compressed sensing image reconstruction method and system based on generation countermeasure network |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811404831.XA CN109544656B (en) | 2018-11-23 | 2018-11-23 | Compressed sensing image reconstruction method and system based on generation countermeasure network |
Publications (2)
Publication Number | Publication Date |
---|---|
CN109544656A true CN109544656A (en) | 2019-03-29 |
CN109544656B CN109544656B (en) | 2023-05-23 |
Family
ID=65850265
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201811404831.XA Active CN109544656B (en) | 2018-11-23 | 2018-11-23 | Compressed sensing image reconstruction method and system based on generation countermeasure network |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109544656B (en) |
Cited By (21)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110211045A (en) * | 2019-05-29 | 2019-09-06 | 电子科技大学 | Super-resolution face image method based on SRGAN network |
CN110361778A (en) * | 2019-07-04 | 2019-10-22 | 华北电力大学 | A kind of Reconstruction of seismic data method based on generation confrontation network |
CN110472667A (en) * | 2019-07-19 | 2019-11-19 | 广东工业大学 | Small object classification method based on deconvolution neural network |
CN110647864A (en) * | 2019-09-30 | 2020-01-03 | 上海依图网络科技有限公司 | Single multi-graph feature recognition method, equipment and medium based on generation countermeasure network |
CN110677671A (en) * | 2019-11-01 | 2020-01-10 | 合肥图鸭信息科技有限公司 | Image compression method and device and terminal equipment |
CN110942097A (en) * | 2019-11-28 | 2020-03-31 | 北京理工大学 | Imaging-free classification method and system based on single-pixel detector |
CN111080727A (en) * | 2019-12-17 | 2020-04-28 | 华中科技大学鄂州工业技术研究院 | Color image reconstruction method and device and image classification method and device |
CN111127454A (en) * | 2019-12-27 | 2020-05-08 | 上海交通大学 | Method and system for generating industrial defect sample based on deep learning |
CN111224992A (en) * | 2020-01-10 | 2020-06-02 | 上海理工大学 | Electric energy quality signal compression reconstruction method based on generation countermeasure network |
CN111382837A (en) * | 2020-02-05 | 2020-07-07 | 鹏城实验室 | Countermeasure sample generation method based on depth product quantization |
CN111461984A (en) * | 2020-03-30 | 2020-07-28 | 华南理工大学 | Load prediction method based on compressed sensing |
CN111711820A (en) * | 2020-05-13 | 2020-09-25 | 哈尔滨工业大学 | Crack image compression sampling method based on generation countermeasure network |
CN111723521A (en) * | 2020-06-12 | 2020-09-29 | 杭州电子科技大学 | Method for recovering limited caliber data into full caliber data based on GAN network |
CN111784581A (en) * | 2020-07-03 | 2020-10-16 | 苏州兴钊防务研究院有限公司 | SAR image super-resolution reconstruction method based on self-normalization generation countermeasure network |
CN112116601A (en) * | 2020-08-18 | 2020-12-22 | 河南大学 | Compressive sensing sampling reconstruction method and system based on linear sampling network and generation countermeasure residual error network |
CN113052925A (en) * | 2021-04-02 | 2021-06-29 | 广东工业大学 | Compressed sensing reconstruction method and system based on deep learning |
CN113361583A (en) * | 2021-06-01 | 2021-09-07 | 珠海大横琴科技发展有限公司 | Countermeasure sample detection method and device |
CN113627538A (en) * | 2021-08-12 | 2021-11-09 | 群联电子股份有限公司 | Method and electronic device for training asymmetric generation countermeasure network to generate image |
CN113808062A (en) * | 2019-04-28 | 2021-12-17 | 深圳市商汤科技有限公司 | Image processing method and device |
CN113808275A (en) * | 2021-09-24 | 2021-12-17 | 南京信息工程大学 | Single-image three-dimensional reconstruction method based on GCN and topology modification |
CN116188259A (en) * | 2022-12-08 | 2023-05-30 | 中国科学院水生生物研究所 | Fish image generation method based on compressed sensing and countermeasure generation |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107784676A (en) * | 2017-09-20 | 2018-03-09 | 中国科学院计算技术研究所 | Compressed sensing calculation matrix optimization method and system based on autocoder network |
CN108171762A (en) * | 2017-12-27 | 2018-06-15 | 河海大学常州校区 | System and method for is reconfigured quickly in a kind of similar image of the compressed sensing of deep learning |
-
2018
- 2018-11-23 CN CN201811404831.XA patent/CN109544656B/en active Active
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107784676A (en) * | 2017-09-20 | 2018-03-09 | 中国科学院计算技术研究所 | Compressed sensing calculation matrix optimization method and system based on autocoder network |
CN108171762A (en) * | 2017-12-27 | 2018-06-15 | 河海大学常州校区 | System and method for is reconfigured quickly in a kind of similar image of the compressed sensing of deep learning |
Cited By (31)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113808062A (en) * | 2019-04-28 | 2021-12-17 | 深圳市商汤科技有限公司 | Image processing method and device |
CN110211045A (en) * | 2019-05-29 | 2019-09-06 | 电子科技大学 | Super-resolution face image method based on SRGAN network |
CN110211045B (en) * | 2019-05-29 | 2022-09-06 | 电子科技大学 | Super-resolution face image reconstruction method based on SRGAN network |
CN110361778A (en) * | 2019-07-04 | 2019-10-22 | 华北电力大学 | A kind of Reconstruction of seismic data method based on generation confrontation network |
CN110472667A (en) * | 2019-07-19 | 2019-11-19 | 广东工业大学 | Small object classification method based on deconvolution neural network |
CN110472667B (en) * | 2019-07-19 | 2024-01-09 | 广东工业大学 | Small target classification method based on deconvolution neural network |
CN110647864A (en) * | 2019-09-30 | 2020-01-03 | 上海依图网络科技有限公司 | Single multi-graph feature recognition method, equipment and medium based on generation countermeasure network |
CN110677671A (en) * | 2019-11-01 | 2020-01-10 | 合肥图鸭信息科技有限公司 | Image compression method and device and terminal equipment |
CN110942097A (en) * | 2019-11-28 | 2020-03-31 | 北京理工大学 | Imaging-free classification method and system based on single-pixel detector |
CN111080727A (en) * | 2019-12-17 | 2020-04-28 | 华中科技大学鄂州工业技术研究院 | Color image reconstruction method and device and image classification method and device |
CN111080727B (en) * | 2019-12-17 | 2023-03-21 | 华中科技大学鄂州工业技术研究院 | Color image reconstruction method and device and image classification method and device |
CN111127454A (en) * | 2019-12-27 | 2020-05-08 | 上海交通大学 | Method and system for generating industrial defect sample based on deep learning |
CN111224992B (en) * | 2020-01-10 | 2022-10-14 | 上海理工大学 | Electric energy quality signal compression reconstruction method based on generation countermeasure network |
CN111224992A (en) * | 2020-01-10 | 2020-06-02 | 上海理工大学 | Electric energy quality signal compression reconstruction method based on generation countermeasure network |
CN111382837A (en) * | 2020-02-05 | 2020-07-07 | 鹏城实验室 | Countermeasure sample generation method based on depth product quantization |
CN111461984B (en) * | 2020-03-30 | 2022-06-14 | 华南理工大学 | Load prediction method based on compressed sensing |
CN111461984A (en) * | 2020-03-30 | 2020-07-28 | 华南理工大学 | Load prediction method based on compressed sensing |
CN111711820A (en) * | 2020-05-13 | 2020-09-25 | 哈尔滨工业大学 | Crack image compression sampling method based on generation countermeasure network |
CN111723521B (en) * | 2020-06-12 | 2023-12-08 | 杭州电子科技大学 | Method for recovering limited caliber data into full caliber data based on GAN network |
CN111723521A (en) * | 2020-06-12 | 2020-09-29 | 杭州电子科技大学 | Method for recovering limited caliber data into full caliber data based on GAN network |
CN111784581A (en) * | 2020-07-03 | 2020-10-16 | 苏州兴钊防务研究院有限公司 | SAR image super-resolution reconstruction method based on self-normalization generation countermeasure network |
CN111784581B (en) * | 2020-07-03 | 2024-07-05 | 苏州兴钊防务研究院有限公司 | SAR image super-resolution reconstruction method based on self-normalization generation countermeasure network |
CN112116601B (en) * | 2020-08-18 | 2023-04-28 | 河南大学 | Compressed sensing sampling reconstruction method and system based on generation of countermeasure residual error network |
CN112116601A (en) * | 2020-08-18 | 2020-12-22 | 河南大学 | Compressive sensing sampling reconstruction method and system based on linear sampling network and generation countermeasure residual error network |
CN113052925A (en) * | 2021-04-02 | 2021-06-29 | 广东工业大学 | Compressed sensing reconstruction method and system based on deep learning |
CN113361583A (en) * | 2021-06-01 | 2021-09-07 | 珠海大横琴科技发展有限公司 | Countermeasure sample detection method and device |
CN113627538A (en) * | 2021-08-12 | 2021-11-09 | 群联电子股份有限公司 | Method and electronic device for training asymmetric generation countermeasure network to generate image |
CN113627538B (en) * | 2021-08-12 | 2024-03-01 | 群联电子股份有限公司 | Method for training asymmetric generation of image generated by countermeasure network and electronic device |
CN113808275A (en) * | 2021-09-24 | 2021-12-17 | 南京信息工程大学 | Single-image three-dimensional reconstruction method based on GCN and topology modification |
CN113808275B (en) * | 2021-09-24 | 2023-10-13 | 南京信息工程大学 | Single image three-dimensional reconstruction method based on GCN and topology modification |
CN116188259A (en) * | 2022-12-08 | 2023-05-30 | 中国科学院水生生物研究所 | Fish image generation method based on compressed sensing and countermeasure generation |
Also Published As
Publication number | Publication date |
---|---|
CN109544656B (en) | 2023-05-23 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109544656A (en) | A kind of compressed sensing image rebuilding method and system based on generation confrontation network | |
CN102144923B (en) | Method and magnetic resonance system to reduce distortions in diffusion imaging | |
CN112116601B (en) | Compressed sensing sampling reconstruction method and system based on generation of countermeasure residual error network | |
CN111383741B (en) | Method, device and equipment for establishing medical imaging model and storage medium | |
CN107341776A (en) | Single frames super resolution ratio reconstruction method based on sparse coding and combinatorial mapping | |
CN114972803B (en) | Snapshot type spectrum imaging method and system based on joint optimization | |
CN111711820A (en) | Crack image compression sampling method based on generation countermeasure network | |
CN113470002B (en) | Chromatography SAR three-dimensional point cloud reconstruction quality evaluation method and device | |
CN110674925A (en) | No-reference VR video quality evaluation method based on 3D convolutional neural network | |
CN102436655A (en) | Super-resolution reconstruction image quality evaluation method based on SVD (singular value decomposition) | |
CN109461122A (en) | A kind of compressed sensing image rebuilding method based on multi-view image | |
CN114140442A (en) | Deep learning sparse angle CT reconstruction method based on frequency domain and image domain degradation perception | |
CN115880158A (en) | Blind image super-resolution reconstruction method and system based on variational self-coding | |
CN111383742A (en) | Method, device, equipment and storage medium for establishing medical imaging model | |
CN112132924B (en) | CT reconstruction method based on deep neural network | |
CN105608719B (en) | A kind of rapid CT image rebuilding method based on two benches projection adjustment | |
CN110926611A (en) | Noise suppression method applied to compressed sensing spectral imaging system | |
CN116228520A (en) | Image compressed sensing reconstruction method and system based on transform generation countermeasure network | |
CN113506362B (en) | Method for synthesizing new view of single-view transparent object based on coding and decoding network | |
CN111241725B (en) | Structure response reconstruction method for generating countermeasure network based on conditions | |
JP2023553004A (en) | Deep learning-based image augmentation for additive manufacturing | |
CN109840544B (en) | Hyperspectral image multi-end member spectrum hybrid analysis method and device | |
CN115115803B (en) | Fetal brain magnetic resonance motion correction method and device based on feature association fusion | |
CN115494439B (en) | Space-time coding image correction method based on deep learning | |
Wasif et al. | Towards a Benchmark EO Semantic Segmentation Dataset for Uncertainty Quantification |
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 | ||
CB02 | Change of applicant information | ||
CB02 | Change of applicant information |
Address after: 210044 No. 219 Ningliu Road, Jiangbei New District, Nanjing City, Jiangsu Province Applicant after: Nanjing University of Information Science and Technology Address before: 210000 Yuting Square, 59 Wangqiao Road, Liuhe District, Nanjing City, Jiangsu Province Applicant before: Nanjing University of Information Science and Technology |
|
GR01 | Patent grant | ||
GR01 | Patent grant |