CN111489305B - Image enhancement method based on reinforcement learning - Google Patents
Image enhancement method based on reinforcement learning Download PDFInfo
- Publication number
- CN111489305B CN111489305B CN202010244525.5A CN202010244525A CN111489305B CN 111489305 B CN111489305 B CN 111489305B CN 202010244525 A CN202010244525 A CN 202010244525A CN 111489305 B CN111489305 B CN 111489305B
- Authority
- CN
- China
- Prior art keywords
- processing
- training
- different degrees
- picture
- different
- 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
- 238000000034 method Methods 0.000 title claims abstract description 18
- 230000002787 reinforcement Effects 0.000 title claims abstract description 11
- 238000012545 processing Methods 0.000 claims abstract description 83
- 238000012549 training Methods 0.000 claims abstract description 27
- 238000012360 testing method Methods 0.000 claims abstract description 17
- 238000007781 pre-processing Methods 0.000 claims abstract description 4
- 230000006870 function Effects 0.000 claims description 19
- 238000011084 recovery Methods 0.000 claims description 13
- 238000013527 convolutional neural network Methods 0.000 claims description 10
- 230000000694 effects Effects 0.000 claims description 10
- 230000001186 cumulative effect Effects 0.000 claims description 6
- 230000006835 compression Effects 0.000 claims description 5
- 238000007906 compression Methods 0.000 claims description 5
- 238000013461 design Methods 0.000 claims description 4
- 238000011056 performance test Methods 0.000 claims description 2
- 238000013528 artificial neural network Methods 0.000 description 3
- 238000004458 analytical method Methods 0.000 description 2
- HPTJABJPZMULFH-UHFFFAOYSA-N 12-[(Cyclohexylcarbamoyl)amino]dodecanoic acid Chemical compound OC(=O)CCCCCCCCCCCNC(=O)NC1CCCCC1 HPTJABJPZMULFH-UHFFFAOYSA-N 0.000 description 1
- 239000000654 additive Substances 0.000 description 1
- 230000000996 additive effect Effects 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 230000004888 barrier function Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 230000015556 catabolic process Effects 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000000354 decomposition reaction Methods 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 238000006731 degradation reaction Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000002708 enhancing effect Effects 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 238000003384 imaging method Methods 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/70—Denoising; Smoothing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/73—Deblurring; Sharpening
-
- 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
- G06T2207/20081—Training; Learning
-
- 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)
- Data Mining & Analysis (AREA)
- Molecular Biology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Image Processing (AREA)
Abstract
The invention relates to an image enhancement method based on reinforcement learning, which comprises the following steps: making a distorted picture data set: preprocessing a training set and a testing set by using matlab by adopting three types of processing modes with different degrees to generate a distortion picture; designing an image enhancement processing tool: respectively training enhancement algorithm parameters aiming at different types or different degrees of distortion pictures, generating corresponding meta files, and obtaining a plurality of processing tools, wherein each tool correspondingly processes distortion of a specific degree and a specific type; training an optimal processing tool to select a network; and (5) testing the performance of the model.
Description
Technical Field
The invention belongs to the technical field of image processing, and relates to a method for improving image quality by reinforcement learning based on a machine learning technology.
Background
With the advent of the big data age, image information is more intuitive and more popular and easy to understand, and is an indispensable information source in daily life. However, the requirement for knowing information by using the picture is to obtain a high-quality picture, and the degraded picture contains a lot of noise, which brings various barriers to subsequent analysis. With the development of artificial intelligence, the requirements of people on picture quality are also increasing. Image enhancement has also been a hotspot in research in the field of computer vision.
In many digital image processing applications, it is often desirable to process and analyze with high quality pictures or video. Because of the limitations of the prior art, imaging devices often obtain low quality pictures, including but not limited to with gaussian white noise, too low resolution, blurring, etc., which can present difficulties for subsequent image processing and analysis, and how to improve the quality of these pictures has become a focus of attention.
In recent years, deep learning, and in particular convolutional neural networks, has proven to be an efficient data-driven framework and has shown good results in underlying image processing problems. Dong et al designed a convolutional neural network architecture [1] to solve the problem of single Zhang Tupian super-resolution reconstruction, and designed three-layer convolutional neural networks respectively simulate sparse coding-based super-resolution reconstruction. Wang et al incorporate neural networks into the sparse coding framework to solve the super-resolution reconstruction problem [2], and skillfully apply this design to the image super-resolution problem by utilizing a neural network design for fast sparse coding. Schulter et al developed a multi-layer perceptron method [3] of removing noise and artifacts by deconvolution construction. Xu et al use a deep convolutional neural network [4] to recover noisy images, using a singular value decomposition approach, reducing parameters in the network. Kim et al [5] found that neural network convergence can be aided by learning the residual between low resolution pictures and high resolution pictures to add more layers to the network.
[1]Dong C,Chen C L,He K,et al.Image super-Resolution Using Deep Convolutional Networks[J].IEEE Transactions onPatternAnalysis&Machine Intelligence,2016,38(2):295-307
[2]Wang Z,Liu D,Yang J,et al.Deep networks for Image Super-Resolution with Sparse Prior[C].In proceedingofICCV,Santiago,Chile,2015:370-378.
[3]Schuler C J,Burger H C,Haemeling S,et al.A Machine Learning Approach for Non-blind Image Deconvolution[C].InProceedingofCVPR,Portland,ORUSA,2013:1067-1074
[4]Xu L,Ren J S,Liu C,et al.Deep convolutional neural network for image deconvolution[C].In Proceeding ofNIPS,Montreal,Quebec,Canada,2014:1790-1798.
[5]Kim J,Lee J K,Lee KM.Accurate Image Super-Resolution Using Very Deep ConvolutionNetworks[C].InProceedings ofCVPR,LasVegas,NV,USA,2016:1646-1654
Disclosure of Invention
The invention aims to provide an image enhancement method based on reinforcement learning, which selects the most suitable image enhancement method, processes the image, enriches the detail information of the image, enhances the image quality, facilitates the subsequent processing and improves the efficiency and the accuracy. The technical proposal is as follows:
an image enhancement method based on reinforcement learning, comprising the following steps:
the first step: making distorted picture datasets
Dividing the disclosed picture data set into a training set and a testing set, preprocessing the training set and the testing set by using matlab by adopting three types of processing modes with different degrees to generate a distorted picture, wherein the distorted picture comprises JPEG compression processing modes with different degrees, gaussian noise processing modes with different degrees and Gaussian blur processing modes with different degrees;
and a second step of: design image enhancement processing tool
Respectively training enhancement algorithm parameters aiming at different types or different degrees of distortion pictures, generating corresponding meta files, and obtaining a plurality of processing tools, wherein each tool correspondingly processes distortion of a specific degree and a specific type, and the enhancement algorithm parameters are recovered by adopting a generated countermeasure network recovery reconstruction algorithm aiming at different degrees of JPEG compression processing modes; aiming at Gaussian noise processing modes with different degrees, recovering by adopting a convolutional neural network denoising algorithm; aiming at Gaussian blur processing modes with different degrees, a convolutional neural network deblurring algorithm is adopted for recovery;
and a third step of: training optimal processing tool selection networks
When reconstructing the distorted image, the recovery effects of different processing tools are different, the recovery effects of different processing sequences are also different, and a network for independently selecting the optimal processing tool is required to be designed; adopting a DQN reinforcement learning algorithm, regarding the selection problem as a Markov process, evaluating each action by using a return function, taking the most suitable action to convert the state to maximize the return function in the face of different current states, regarding the selection of processing tools as discrete actions, and selecting the most suitable processing tools and processing sequences in the face of distortion pictures of different degrees;
determining a training optimal processing tool to select a network parameter adjustment scheme: batch is set to 32, learning rate is set to 0.0001, initial value of exploration rate is set to 0.1, and iteration times are set to 100000 times; setting Batch to 1 during testing, namely only processing one image at a time; training an optimal processing tool to select network parameters so as to maximize an objective function accumulated return function; when the training iteration is finished or the cumulative return function converges, the optimal processing tool for obtaining the distorted picture selects a network: inputting a distorted picture, and outputting a label and a processing sequence corresponding to an optimal processing tool;
fourth step: model performance test
Inputting the distorted pictures in the test set into an optimal processing tool selection network to obtain the corresponding labels and processing sequences of the picture processing tools, and performing corresponding operation on the distorted pictures to obtain enhanced pictures; and the performance of the model is evaluated by calculating the peak power signal-to-noise ratio PSNR between the undistorted original picture and the enhanced image, and the higher the PSNR value is, the better the recovery effect is.
The beneficial effects of the invention are shown in table 1.
Table 1 results statistics table
Drawings
FIG. 1 image enhancement model Structure
FIG. 2JPEG compressed picture
FIG. 3 compressed image model enhancement results graph
FIG. 4 Gaussian noise plot
FIG. 5 noise image model enhancement results graph
FIG. 6 is a blurring map
FIG. 7 is a fuzzy image model enhanced result diagram
Detailed Description
In order to make the technical scheme of the invention clearer, the following detailed description of the invention is further described with reference to the accompanying drawings.
The first step: a distorted picture dataset is produced.
The DIV2K dataset was divided into training sets and test sets at a ratio of 15:1. Preprocessing the training set and the test set by using matlab in 12 processing modes to generate distorted pictures, and obtaining 750 training set pictures and 50 test set pictures, wherein each picture is subjected to multiple processing. The 12 modes used in the present invention are shown in table 2.
TABLE 2 distortion handling
(1) Gaussian blur processing
| 1/N | ... | 1/N |
| ... | 1/N | ... |
| 1/N | ... | 1/N |
FIG. 1 [ N, N ] Gaussian blur convolution kernel
And performing traversal convolution on the image by using the convolution check image shown in fig. 1 to obtain a picture after blurring processing.
(2) Adding Gaussian noise
An input image f (x, y) is processed to produce a degraded image g (x, y). Given g (x, y), the degradation function H, and the additive noise term η (x, y), the degraded image in the spatial domain can be given by:
g(x,y)＝h(x,y)*f(x,y)+η(x,y)
in the frequency domain:
G(u,v)＝H(u,v)F(u,v)+N(u,v)
and a second step of: an image enhancement processing tool is designed.
And respectively training the enhancement algorithm parameters aiming at different types or different degrees of distortion, and generating corresponding meta files to obtain 12 processing tools for processing the distortion with specific degree and specific type. In the invention, aiming at 4 different degrees of JPEG compression, adopting a generated countermeasure network restoration reconstruction algorithm for restoration; aiming at 4 Gaussian noises with different degrees, recovering by adopting a convolutional neural network denoising algorithm; and (4) for 4 kinds of blurring at different degrees, recovering by adopting a convolutional neural network deblurring algorithm.
And a third step of: the optimal processing tool is trained to select a network.
When reconstructing a distorted image, the recovery effects of different processing tools are different, and the recovery effects of different processing sequences are also different, so that a network for autonomously selecting an optimal processing tool should be designed. Reinforcement learning algorithms, consider the selection problem as a markov process, with a return function to evaluate each action. The most appropriate action is taken to switch states to maximize the reward function in the face of different current states. The present invention adopts DQN reinforcement learning algorithm to consider the selection of processing tools as discrete actions, and selects the most appropriate processing tools and processing sequences for distorted pictures of different degrees, as shown in fig. 2.
In the DQN algorithm of the present invention, the environmental state S t ＝{I t ,v t }, wherein I t Representing an input distorted picture vector, v t Representing a historical motion vector, v in the first step t Is a 0 vector; action A taken by the individual at time t t E {12 tools }, taking action, i.e. selecting one tool to process the distorted picture, obtaining the reconstructed picture, and converting the state to S t+1 At the same time get the rewards R of the environment t The calculation formula is as follows:
R t ＝||I target -I t-1 || 2 -||I target -I t || 2
wherein I is target Representing undistorted artwork; cumulative return function Q (t) =e (R t+1 +λR t+2 +λ 2 R t+3 +…|S t ) Where E is the desired function, λ is the decay factor, and the cumulative return function Q (t) is maximized as equivalent to selecting the optimal processing tool problem.
Because the training pictures used by the invention are larger in size, the training optimal processing tool is determined to select a network parameter adjustment scheme through multiple experimental results: batch is set to 32, learning rate is set to 0.0001, initial value of exploration rate is set to 0.1, and iteration number is set to 100000. The test sets Batch to 1, i.e., only one image is processed at a time. The training optimization processing tool selects network parameters such that the objective function cumulative return function Q (t) is maximized. The experimental environment is Ubuntu16.04 operating system, and the training is performed by using RTX2060GPU of 6GB video memory of NVIDIA company and the training is accelerated by using CUDA.
When the training iteration is finished or the cumulative return function Q (t) converges, a distortion picture optimal processing tool selection network is obtained: and inputting the distorted pictures, and outputting the corresponding labels and the processing sequence of the optimal processing tool.
Fourth step: and (5) testing the performance of the model.
Inputting the distorted pictures in the test set into an optimal processing tool selection network to obtain the corresponding labels and processing sequences of the picture processing tools, and carrying out corresponding operation on the distorted pictures to obtain the enhanced pictures. And the performance of the model is evaluated by calculating the peak power signal-to-noise ratio PSNR between the undistorted original picture and the enhanced image, and the higher the PSNR value is, the better the recovery effect is.
PSNR is defined as:
where m, n, c represent the size of the image, 256,8 in the present invention; x is an undistorted original picture, y is a reconstructed picture, MAX I Is the pixel maximum, i.e., 255.
And analyzing and processing the experimental data to evaluate the picture quality enhancement performance of the invention. The results after the test are shown in table 1, and comparison shows that the invention has better effect on enhancing the picture quality.
Claims (1)
1. An image enhancement method based on reinforcement learning, comprising the following steps:
the first step: making distorted picture datasets
Dividing the disclosed picture data set into a training set and a testing set, preprocessing the training set and the testing set by using matlab by adopting three types of processing modes with different degrees to generate a distorted picture, wherein the distorted picture comprises JPEG compression processing modes with different degrees, gaussian noise processing modes with different degrees and Gaussian blur processing modes with different degrees;
and a second step of: design image enhancement processing tool
Respectively training enhancement algorithm parameters aiming at different types or different degrees of distortion pictures, generating corresponding meta files, and obtaining a plurality of processing tools, wherein each tool correspondingly processes distortion of a specific degree and a specific type, and the enhancement algorithm parameters are recovered by adopting a generated countermeasure network recovery reconstruction algorithm aiming at different degrees of JPEG compression processing modes; aiming at Gaussian noise processing modes with different degrees, recovering by adopting a convolutional neural network denoising algorithm; aiming at Gaussian blur processing modes with different degrees, a convolutional neural network deblurring algorithm is adopted for recovery;
and a third step of: training optimal processing tool selection networks
When reconstructing the distorted image, the recovery effects of different processing tools are different, the recovery effects of different processing sequences are also different, and a network for independently selecting the optimal processing tool is required to be designed; adopting a DQN reinforcement learning algorithm, regarding the selection problem as a Markov process, evaluating each action by using a return function, taking the most suitable action to convert the state to maximize the return function in the face of different current states, regarding the selection of processing tools as discrete actions, and selecting the most suitable processing tools and processing sequences in the face of distortion pictures of different degrees;
determining a training optimal processing tool to select a network parameter adjustment scheme: batch is set to 32, learning rate is set to 0.0001, initial value of exploration rate is set to 0.1, and iteration times are set to 100000 times; setting Batch to 1 during testing, namely only processing one image at a time; training an optimal processing tool to select network parameters so as to maximize an objective function accumulated return function; when the training iteration is finished or the cumulative return function converges, the optimal processing tool for obtaining the distorted picture selects a network: inputting a distorted picture, and outputting a label and a processing sequence corresponding to an optimal processing tool;
fourth step: model performance test
Inputting the distorted pictures in the test set into an optimal processing tool selection network to obtain the corresponding labels and processing sequences of the picture processing tools, and performing corresponding operation on the distorted pictures to obtain enhanced pictures; and the performance of the model is evaluated by calculating the peak power signal-to-noise ratio PSNR between the undistorted original picture and the enhanced image, and the higher the PSNR value is, the better the recovery effect is.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010244525.5A CN111489305B (en) | 2020-03-31 | 2020-03-31 | Image enhancement method based on reinforcement learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010244525.5A CN111489305B (en) | 2020-03-31 | 2020-03-31 | Image enhancement method based on reinforcement learning |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111489305A CN111489305A (en) | 2020-08-04 |
| CN111489305B true CN111489305B (en) | 2023-05-30 |
Family
ID=71794572
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010244525.5A Active CN111489305B (en) | 2020-03-31 | 2020-03-31 | Image enhancement method based on reinforcement learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111489305B (en) |
Families Citing this family (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112529064B (en) * | 2020-12-03 | 2022-09-02 | 燕山大学 | Efficient real-time semantic segmentation method |
| CN112365428B (en) * | 2020-12-03 | 2022-04-01 | 华中科技大学 | DQN-based highway monitoring video defogging method and system |
| CN113139578B (en) * | 2021-03-23 | 2022-12-06 | 广东省科学院智能制造研究所 | Deep learning image classification method and system based on optimal training set |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107770517A (en) * | 2017-10-24 | 2018-03-06 | 天津大学 | Full reference image quality appraisement method based on image fault type |
| CN110598018A (en) * | 2019-08-13 | 2019-12-20 | 天津大学 | Sketch image retrieval method based on cooperative attention |
| CN110866872A (en) * | 2019-10-10 | 2020-03-06 | 北京邮电大学 | Pavement crack image preprocessing intelligent selection method and device and electronic equipment |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2018053340A1 (en) * | 2016-09-15 | 2018-03-22 | Twitter, Inc. | Super resolution using a generative adversarial network |
-
2020
- 2020-03-31 CN CN202010244525.5A patent/CN111489305B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107770517A (en) * | 2017-10-24 | 2018-03-06 | 天津大学 | Full reference image quality appraisement method based on image fault type |
| CN110598018A (en) * | 2019-08-13 | 2019-12-20 | 天津大学 | Sketch image retrieval method based on cooperative attention |
| CN110866872A (en) * | 2019-10-10 | 2020-03-06 | 北京邮电大学 | Pavement crack image preprocessing intelligent selection method and device and electronic equipment |
Non-Patent Citations (2)
| Title |
|---|
| Ke Yu等."Crafting a Toolchain for Image Restoration by Deep Reinforcement Learning".arXiv.2018,第1-10页. * |
| 谢宁 ; 赵婷婷 ; 杨阳 ; 魏琴 ; Heng Tao SHEN.基于创意序列学习的艺术风格学习与绘制系统.软件学报.2017,(第004期),全文. * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111489305A (en) | 2020-08-04 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112884851B (en) | Construction method of deep compressed sensing network based on expansion iteration optimization algorithm | |
| CN111489305B (en) | Image enhancement method based on reinforcement learning | |
| CN109410146A (en) | A kind of image deblurring algorithm based on Bi-Skip-Net | |
| CN112288632B (en) | Single image super-resolution method and system based on simplified ESRGAN | |
| CN110796622B (en) | Image bit enhancement method based on multi-layer characteristics of series neural network | |
| CN105631807A (en) | Single-frame image super resolution reconstruction method based on sparse domain selection | |
| CN105590296B (en) | A kind of single-frame images Super-Resolution method based on doubledictionary study | |
| CN116486074A (en) | Medical image segmentation method based on local and global context information coding | |
| CN116188272B (en) | Two-stage depth network image super-resolution reconstruction method suitable for multiple fuzzy cores | |
| Guo et al. | Agem: Solving linear inverse problems via deep priors and sampling | |
| CN114757828A (en) | Transformer-based video space-time super-resolution method | |
| Peng et al. | Progressive training of multi-level wavelet residual networks for image denoising | |
| Jang et al. | Dual path denoising network for real photographic noise | |
| CN114742911A (en) | Image compressed sensing reconstruction method, system, equipment and medium | |
| Wu et al. | Feedback pyramid attention networks for single image super-resolution | |
| CN110288529B (en) | Single image super-resolution reconstruction method based on recursive local synthesis network | |
| CN114897704A (en) | Single-image super-resolution algorithm based on feedback mechanism | |
| Amaranageswarao et al. | Residual learning based densely connected deep dilated network for joint deblocking and super resolution | |
| Jaisurya et al. | Attention-based single image dehazing using improved cyclegan | |
| CN111489306A (en) | Image denoising method based on reinforcement learning | |
| CN116485651A (en) | Image super-resolution reconstruction method | |
| Zhang et al. | Image super-resolution via RL-CSC: when residual learning meets convolutional sparse coding | |
| Hu et al. | MAUN: Memory-Augmented Deep Unfolding Network for Hyperspectral Image Reconstruction | |
| Xue et al. | MMPDNet: Multi-Stage & Multi-Attention Progressive Image Denoising | |
| CN112308772A (en) | Super-resolution reconstruction method based on deep learning local and non-local information |
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 |