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 PDF

Info

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
image
network
original 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
Application number
CN201811404831.XA
Other languages
Chinese (zh)
Other versions
CN109544656B (en
Inventor
孙玉宝
陈基伟
刘青山
徐宏伟
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Nanjing University of Information Science and Technology
Original Assignee
Nanjing University of Information Science and Technology
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Nanjing University of Information Science and Technology filed Critical Nanjing University of Information Science and Technology
Priority to CN201811404831.XA priority Critical patent/CN109544656B/en
Publication of CN109544656A publication Critical patent/CN109544656A/en
Application granted granted Critical
Publication of CN109544656B publication Critical patent/CN109544656B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

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

A kind of compressed sensing image rebuilding method and system based on generation confrontation network
Technical field
The present invention relates to technical field of image information processing, and in particular to a kind of based on the compressed sensing for generating confrontation network Image rebuilding method and system.
Background technique
Compressed sensing (Compressed Sensing, CS) is that a kind of novel signal acquisition is theoretical, has been merged traditional Sampling and compression process, can directly acquire the measurement data far below Nyquist sample rate, can reduce sampling cost, reduce Storage resource, while the coding side of compressed sensing model need to only carry out linear random measurement, and the complex optimization of reconstruction signal Process is completed in decoding end.
Those imaging systems are based on compressive sensing theory, use iteration optimization algorithms reconstruct image to a small amount of measured value of observation Picture.However these restructing algorithms require to carry out complicated interative computation, reconstitution time is longer, and in lower sample rate Under, reconstructed image quality is poor, hinders the deep development and industrial application of compressed sensing imaging.
Summary of the invention
Goal of the invention: for overcome the deficiencies in the prior art, the present invention provides a kind of based on the compression for generating confrontation network Perceptual image method for reconstructing, this method can solve that reconstitution time is longer, the poor problem of reconstructed image quality, and the present invention is also It provides a kind of based on the compressed sensing image re-construction system for generating confrontation network.
Technical solution: of the present invention based on the compressed sensing image rebuilding method for generating confrontation network, comprising:
S1, the measurement vector obtained according to image sampling and reconstruction image size construct generation neural network based and fight Network model, and designed for optimizing the objective function for generating 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, Model output is the correspondence original image by the measurement vector reconstruction out；Otherwise S2-S4 is returned to step.
Preferably, in the S1, measurement vector is expressed as:
Y=Φ x+ ξ, y ∈ RM, Φ ∈ RM×N, x ∈ RN
Wherein, y indicates measurement vector, and Φ is calculation matrix, and x indicates that the data matrix of the image to be sampled becomes vector The data row vector become after change, M indicate measurement vector magnitude, and N is the pixel number of the image to be sampled.
Preferably, in the S1, the formula of confrontation network model is generated are as follows:
J (D, G)=minG maxD[lrec(G(y))+lreg(G(y))+lGAN(G(y),D)]
Wherein, G is generator, and D is discriminator, and y is measurement vector, and x indicates that the data matrix of the image to be sampled becomes At the data row vector become after vectorization, lrecFor reconstruct loss, lregFor total variation canonical, lGANFor pair of discriminator network Damage-retardation is lost.
Preferably, corresponding formula is lost in the confrontation of the reconstruct loss, total variation canonical and discriminator network to be updated to In the formula for generating confrontation network model, the objective function of optimization generator and discriminator parameter is respectively obtained；
The objective function of the generator indicates are as follows:
The objective function of the discriminator are as follows:
Wherein,For generator reconstructed image, the location of pixels of i and j for original image, pdataAnd pyIt is original graph The statistical distribution of picture and random measurement vector, γ are the loss weight of discriminator.
Preferably, in the S3, the activation primitive used when training the generator is Selu, when training the discriminator The activation primitive used is Lrelu.
Invention additionally discloses a kind of system realized based on the compressed sensing image rebuilding method for generating confrontation network, packets It includes:
Network model constructs module, and the measurement vector and reconstruction image size for being sampled according to described image construct Generation neural network based fights network model, and designed for optimizing the target letter for generating confrontation network model parameter Number；
Parameter presetting module, for the default training parameter generated when fighting network model；
Generator training module, for according to the objective function, using back-propagation algorithm alternately training generator with Discriminator；
Judgment module is restrained, for judging whether the generation confrontation network model restrains, if the generation fights network Model convergence, then trained network can be directly realized by compressed sensing task, and model output is by the measurement vector reconstruction Correspondence original image out；Otherwise the parameter presetting module is returned.
Preferably, in the network model building module,
The measurement vector is expressed as:
Y=Φ x+ ξ, y ∈ RM, Φ ∈ RM×N, x ∈ RN
Wherein, y indicates measurement vector, and Φ is calculation matrix, and x indicates that the data matrix of the image to be sampled becomes vector The data row vector become after change, M indicate measurement vector magnitude, and N is the pixel number of the image to be sampled.
Preferably, in the network model building module, the formula of confrontation network model is generated are as follows:
J (D, G)=minG maxD[lrec(G(y))+lreg(G(y))+lGAN(G(y),D)]
Wherein, G is generator, and D is discriminator, and y is measurement vector, and x indicates that the data matrix of the original image becomes The data row vector become after vectorization, lrecFor reconstruct loss, lregFor total variation canonical, lGANFor the confrontation of discriminator network Loss.
Preferably, corresponding formula is lost in the confrontation of the reconstruct loss, total variation canonical and discriminator network to be updated to In the formula for generating confrontation network model, the objective function of optimization generator and discriminator parameter is respectively obtained；
The objective function of the generator indicates are as follows:
The objective function of the discriminator are as follows:
Wherein,For generator reconstructed image, the location of pixels of i and j for original image, pdataAnd pyIt is original graph The statistical distribution of picture and random measurement vector, γ are the loss weight of discriminator.
Preferably, in the generator training module, the activation primitive used when training the generator is Selu, training The activation primitive used when the discriminator is Lrelu.
The utility model has the advantages that the present invention utilizes compressive sensing theory, establish a kind of based on the compressed sensing figure for generating confrontation network As method for reconstructing, model utilizes the powerful mapping ability of generator, preliminary reconstruction original graph in the case where image low sampling rate Picture, meanwhile, using the dual training of generator and discriminator, the ability of generator is further enhanced, so that generator was rebuild Image pixel is distributed closer to original image, has achieved the purpose that Exact Reconstruction original image under low sampling rate.
Detailed description of the invention
Fig. 1 is flow chart of the method for the present invention；
Fig. 2 is polyalgorithm of the present invention for the reconstruction image of CelebA human face data collection and the comparison knot of original image Fruit, wherein Fig. 2 a is the comparing result schematic diagram of reconstruction image and original image that ReconNet algorithm obtains, and Fig. 2 b is CSGAN calculation The comparing result of reconstruction image and original image that method obtains, Fig. 2 c are the reconstruction image obtained using DCGAN algorithm and original image Comparing result schematic diagram, Fig. 2 d are the comparing result figure of reconstruction image and original image that method of the present invention obtains；
Fig. 3 be it is of the invention involved in ReconNet algorithm, CSGAN algorithm, DCGAN algorithm and this algorithm obtain The mean square error comparing result figure of reconstruction image and original image；
Fig. 4 be it is of the invention involved in ReconNet algorithm, CSGAN algorithm, DCGAN algorithm and this algorithm obtain The comparing result figure of reconstruction image and original image on Y-PSNR and picture structure similarity indices；
Fig. 5 be it is of the invention involved in ReconNet algorithm, CSGAN algorithm, DCGAN algorithm and this algorithm rebuild one Open time comparing result figure required for image；
Fig. 6 is system structure diagram of the present invention.
Specific embodiment
As shown in Figure 1, the present invention provides a kind of based on the compressed sensing image rebuilding method for generating confrontation network, packet It includes:
Step 1 samples original image, obtains measurement vector
Measure vector y=Φ x+ ξ, y ∈ RM, Φ ∈ RM×N, x ∈ RN, y expression measurement vector, Φ is measurement vector, x table Diagram sheet data matrix becomes the data row vector become after vectorization, and M indicates measurement vector magnitude, and N is the picture of original image Prime number.
Step 2 generates confrontation network model according to measurement vector magnitude and the building of reconstruction image size, and designed for excellent Change the objective function of network model parameter.
Wherein, measurement vector magnitude determines that the size of generator input, reconstruction image size determine generator final output Size.The generation fights network model are as follows:
J (D, G)=minG maxD[lrec(G(y))+lreg(G(y))+lGAN(G(y),D)]
In formula, G is generator, and D is discriminator, and reconstruct loss isTotal variation canonical isThe confrontation of discriminator network is lostThe confrontation of reconstruct loss, total variation canonical and discriminator network is lost and is substituted into Above-mentioned network model obtains the objective function that generator and discriminator parameter is separately optimized, the respectively objective function of generator ForThe target of discriminator Function is For generator reconstructed image, i and j are the picture of original image Plain position, pdataAnd pyIt is the statistical distribution of original image and random measurement vector, γ is the loss weight of discriminator.
Step 3, the default super ginseng of network model training
Model learning rate α, the number of iterations L when preset parameter includes trained, training batch size S, training generator When discriminator loss weight γ, the depth and the number of plies of generator and discriminator network, the activation primitive and discriminator of generator Generator classification.
Step 4, according to above-mentioned objective function, using back-propagation algorithm alternately training generator and discriminator.
Specifically, step 4 specifically includes:
Step 41 chooses S training image { x in training set(1),…,x(s), and obtain measuring vector { y accordingly(1),…,y(s)}；
Step 42 passes through back-propagation algorithm, updates the parameter ω in discriminator.
The first step calculates discriminator loss
Second step, fixed generator parameter θ, updates discriminator parameter ω ← ω+α RMSProp (ω, dω).Wherein, RMSProp is one of gradient descent algorithm.
Required generator loss function when step 43, foundation model construction training generator networkCalculate S training in batch The loss function of image, fixed discriminator parameter ω update generator parameter θ: θ ← θ-using RMSProp gradient descent algorithm α·RMSProp(θ,gθ)。
Step 44, all images to entire training set successively carry out step 41, step 42, step 43, carry out L in total Secondary iteration.
If step 5, network model are restrained, the compressed sensing that trained network can then be directly realized by section opposite end is appointed Business, the output of model is the correspondence original image reconstructed by measurement vector, otherwise returns to step 3, and circulation step 3- step 5.
Specifically, the step 5 specifically includes:
Step 51 judges whether generator network and discriminator network restrain.In network training iterative process, work as identification Device loses dωWith generator gθReduce and progressive when some value respectively, determines network for convergence.
The measurement vector obtained after calculation matrix Φ sampling is input to convergent generator network by step 52 Afterwards, the output of generator is corresponding reconstruction image.
If step 53, this time repetitive exercise is not restrained, 3 are returned to step.
To verify effect of the invention, emulation experiment is carried out to the present invention, test image specification is 64 × 64, in CelebA Trained and test model on human face data collection, setting relevant parameter: model learning rate α=0.002, the number of iterations L=30, every time The network of the picture number S=16 of input, γ=0.01, generator and discriminator from be input to the process of output in the following table 1 and In table 2.
The training depth and activation primitive of 1 generator of table
The training depth and activation primitive of 2 discriminator of table
The evaluation of experiment uses qualitative and quantitative two kinds of analysis methods.
It is respectively 20,100 in sampling number of blocks that Fig. 2, which gives the present invention and tri- kinds of algorithms of ReconNet, CSGAN, DCGAN, Image reconstruction Contrast on effect when with 500.Sampling block is the size for measuring vector y herein, from figure 2 it can be seen that for Same four facial images, reconstruction effect of the invention are substantially better than these three algorithms of ReconNet, CSGAN, DCGAN.
Compare about quantitative analysis, carries out judge picture quality using MSE, PSNR and SSIM, use single image Reconstruction time (chronomere is millisecond) judges the reconstruction speed of algorithm.Wherein, MSE is mean square error (Mean Squared Error) i.e. in image single pixel mean error, PSNR be Y-PSNR (Peak Signal to Noise Ratio), SSIM is structural similarity (structural similarity index), is calculated as follows respectively:
Wherein, range indicates the dynamic range of image pixel value, and μ is mean value, and σ is variance, c1=(k1L)2,c1=(k1L)2 It is for maintaining stable constant, L is the dynamic range of pixel value, k1=0.01, k2=0.03.
When making quantitative comparison, we choose 4 test pictures from test set, after being sampled respectively using calculation matrix Φ, It is input in each model, output reconstruction image is calculated by model, reconstruction image is compared with corresponding original image, calculates correspondence MSE, PSNR and SSIM value.
Fig. 3 gives inventive algorithm and ReconNet, CSGAN, DCGAN, TWLST, LASSO these types algorithm exist respectively MSE (under log2 scaling) fiducial value in the test set of CelebA, abscissa are Number of Measurements (sampling Quantity), ordinate is error, that is, mean square error (Reconstruction error per pixel) of every pixel；Fig. 4 is provided Fiducial value of the PSNR and SSIM of inventive algorithm and ReconNet, CSGAN, DCGAN under the different sampling blocks.Fig. 5 is provided This algorithm and ReconNet, CSGAN, DCGAN, TwIST and LASSO these types algorithm respectively mnist, fmnist with And the average time (unit millisecond) of reconstruction piece image compares figure in CelebA test set.
The present invention also provides a kind of based on the compressed sensing image re-construction system for generating confrontation network, as shown in fig. 6, packet It includes:
Network model constructs module, and the measurement vector and reconstruction image size for being sampled according to described image construct Generation neural network based fights network model, and designed for optimizing the target letter for generating confrontation network model parameter Number；
Parameter presetting module, for the default training parameter generated when fighting network model；
Generator training module, for according to the objective function, using back-propagation algorithm alternately training generator with Discriminator；
Judgment module is restrained, for judging whether the generation confrontation network model restrains, if the generation fights network Model convergence, then trained network can be directly realized by compressed sensing task, and model output is by the measurement vector reconstruction Correspondence original image out；Otherwise the parameter presetting module is returned to.
System of the present invention is realized based on the compressed sensing image rebuilding method for generating confrontation network, specific skill Art is similar with method, and details are not described herein by the present invention.The foregoing is only a preferred embodiment of the present invention, but this hair Bright protection scope is not limited thereto, anyone skilled in the art the invention discloses technical scope Interior, any changes or substitutions that can be easily thought of, should be covered by the protection scope of the present invention
In summary, the reconstruct iteration time relative to conventional compression perception is long, and reconstruction quality difference asks under low sampling rate Topic.The present invention utilizes deep learning, establishes a kind of based on the compressed sensing image reconstruction model for generating confrontation network, allows generator G reconstruction image, and be put to improve the pixel distribution of reconstruction image using discriminator D, reconstruction errors are reduced, this makes low sampling rate Under the reconstruction of complicated image have better effect, methods herein has preferable weight for increasingly complex color image Effect is built, either from the error of reconstruction, similarity and efficiency or from visual effect, there is certain advantage.

Claims (10)

1. a kind of based on the compressed sensing image rebuilding method for generating confrontation network characterized by comprising
S1, the measurement vector sampled according to original image and reconstruction image size construct generation neural network based and fight Network model, and designed for optimizing the objective function for generating 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, model Output is the correspondence original image by the measurement vector reconstruction out；Otherwise S2-S4 is returned to step.
2. according to claim 1 based on the compressed sensing image rebuilding method for generating confrontation network, which is characterized in that institute It states in S1, measurement vector is expressed as:
Y=Φ x+ ξ, y ∈ RM, Φ ∈ RM×N, x ∈ RN
Wherein, y indicates measurement vector, and Φ is calculation matrix, and x indicates to become after the data matrix of the original image becomes vectorization At data row vector, M indicate measurement vector magnitude, N be the original image pixel number.
3. according to claim 1 based on the compressed sensing image rebuilding method for generating confrontation network, which is characterized in that institute It states in S1, generates the formula of confrontation network model are as follows:
J (D, G)=minG maxD[lrec(G(y))+lreg(G(y))+lGAN(G(y),D)]
Wherein, G is generator, and D is discriminator, and y is measurement vector, and x indicates that the data matrix of the original image becomes vector The data row vector become after change, lrecFor reconstruct loss, lregFor total variation canonical, lGANIt is lost for the confrontation of discriminator network.
4. according to claim 3 based on the compressed sensing image rebuilding method for generating confrontation network, which is characterized in that will The confrontation of the reconstruct loss, total variation canonical and discriminator network loses corresponding formula and is updated to the generation confrontation network mould In the formula of type, the objective function of optimization generator and discriminator parameter is respectively obtained；
The objective function of the generator indicates are as follows:
The objective function of the discriminator are as follows:
Wherein,For generator reconstructed image, the location of pixels of i and j for original image, pdataAnd pyBe original image and The statistical distribution of random measurement vector, γ are the loss weight of discriminator.
5. according to claim 1 based on the compressed sensing image rebuilding method for generating confrontation network, which is characterized in that institute State in S3, when the training generator activation primitive that uses for Selu, activation primitive that when trained discriminator uses for Lrelu。
6. a kind of be based on what the compressed sensing image rebuilding method for generating confrontation network was realized described in -5 according to claim 1 System characterized by comprising
Network model constructs module, and the measurement vector for being sampled according to original image is based on the building of reconstruction image size The generation of neural network fights network model, and designed for optimizing the objective function for generating confrontation network model parameter；
Parameter presetting module, for the default training parameter generated when fighting network model；
Generator training module, for according to the objective function, alternately training generator using back-propagation algorithm and identifying Device；
Judgment module is restrained, for judging whether the generation confrontation network model restrains, if the generation fights network model Convergence, then trained network can be directly realized by compressed sensing task, and model output is to be gone out by the measurement vector reconstruction Corresponding original image；Otherwise the parameter presetting module is returned.
7. according to claim 6 based on the compressed sensing image re-construction system for generating confrontation network, which is characterized in that institute It states in network model building module,
The measurement vector is expressed as:
Y=Φ x+ ξ, y ∈ RM, Φ ∈ RM×N, x ∈ RN
Wherein, y indicates measurement vector, and Φ is calculation matrix, and x indicates to become after the data matrix of the original image becomes vectorization At data row vector, M indicate measurement vector size, N be the image to be sampled pixel number.
8. according to claim 6 based on the compressed sensing image re-construction system for generating confrontation network, which is characterized in that institute It states in network model building module, generates the formula of confrontation network model are as follows:
J (D, G)=minG maxD[lrec(G(y))+lreg(G(y))+lGAN(G(y),D)]
Wherein, G is generator, and D is discriminator, and y is measurement vector, and x indicates that the data matrix of the original image becomes vector The data row vector become after change, lrecFor reconstruct loss, lregFor total variation canonical, lGANIt is lost for the confrontation of discriminator network.
9. according to claim 8 based on the compressed sensing image re-construction system for generating confrontation network, which is characterized in that will The confrontation of the reconstruct loss, total variation canonical and discriminator network loses corresponding formula and is updated to the generation confrontation network mould In the formula of type, the objective function of optimization generator and discriminator parameter is respectively obtained；
The objective function of the generator indicates are as follows:
The objective function of the discriminator are as follows:
Wherein,For generator reconstructed image, the location of pixels of i and j for original image, pdataAnd pyBe original image and The statistical distribution of random measurement vector, γ are the loss weight of discriminator.
10. according to claim 6 based on the compressed sensing image re-construction system for generating confrontation network, which is characterized in that In the generator training module, the activation primitive used when training the generator is adopted when training the discriminator for Selu Activation primitive is Lrelu.
CN201811404831.XA 2018-11-23 2018-11-23 Compressed sensing image reconstruction method and system based on generation countermeasure network Active CN109544656B (en)

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)

* Cited by examiner, † Cited by third party
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
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
CN113808062A (en) * 2019-04-28 2021-12-17 深圳市商汤科技有限公司 Image processing method and device
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)

* Cited by examiner, † Cited by third party
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

Patent Citations (2)

* Cited by examiner, † Cited by third party
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)

* Cited by examiner, † Cited by third party
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
CN108416840A (en) A kind of dense method for reconstructing of three-dimensional scenic based on monocular camera
CN107451956B (en) A kind of reconstructing method of code aperture spectrum imaging system
CN112116601B (en) Compressed sensing sampling reconstruction method and system based on generation of countermeasure residual error network
CN107525588A (en) A kind of method for fast reconstruction of the double camera spectrum imaging system based on GPU
CN108109147A (en) A kind of reference-free quality evaluation method of blurred picture
CN101950422B (en) Singular value decomposition(SVD)-based image quality evaluation method
CN108038503B (en) Woven fabric texture characterization method based on K-SVD learning dictionary
WO2021248966A1 (en) Point cloud quality assessment method, encoder, decoder, and storage medium
CN106447647A (en) No-reference quality evaluation method of compression perception recovery images
CN110516716A (en) Non-reference picture quality appraisement method based on multiple-limb similarity network
CN112365554A (en) Compressed sensing image reconstruction method based on multi-scale residual error neural network
CN113129272A (en) Defect detection method and device based on denoising convolution self-encoder
CN113177937B (en) Improved YOLOv 4-tiny-based cloth defect detection method
CN110674925B (en) No-reference VR video quality evaluation method based on 3D convolutional neural network
CN110532928A (en) Facial critical point detection method based on facial area standardization and deformable hourglass network
CN114972332B (en) Bamboo laminated wood crack detection method based on image super-resolution reconstruction network
CN104734724B (en) Based on the Compression of hyperspectral images cognitive method for weighting Laplce's sparse prior again
CN111882516B (en) Image quality evaluation method based on visual saliency and deep neural network
CN112364568A (en) Oven food volume and mass estimation algorithm based on computer vision
CN108537104A (en) Compressed sensing network based on full figure observation and perception loss reconstructing method
CN107862683A (en) A kind of more exposure high-dynamics images of synthesis rebuild effect evaluation method
CN117422619A (en) Training method of image reconstruction model, image reconstruction method, device and equipment
CN107220998A (en) The method and system of object temperature are judged using color of image
CN109168002B (en) Video signal measurement domain estimation method based on compressed sensing and convolutional neural network

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