CN114170107A - Turbid underwater polarization image restoration method based on generation countermeasure network - Google Patents
Turbid underwater polarization image restoration method based on generation countermeasure network Download PDFInfo
- Publication number
- CN114170107A CN114170107A CN202111514937.7A CN202111514937A CN114170107A CN 114170107 A CN114170107 A CN 114170107A CN 202111514937 A CN202111514937 A CN 202111514937A CN 114170107 A CN114170107 A CN 114170107A
- Authority
- CN
- China
- Prior art keywords
- network
- image
- underwater
- polarization
- generation
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 230000010287 polarization Effects 0.000 title claims abstract description 47
- 238000000034 method Methods 0.000 title claims abstract description 31
- 238000012549 training Methods 0.000 claims abstract description 15
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 claims abstract description 15
- 238000003384 imaging method Methods 0.000 claims abstract description 11
- 238000012360 testing method Methods 0.000 claims abstract description 4
- 238000012795 verification Methods 0.000 claims abstract description 4
- 230000006870 function Effects 0.000 claims description 18
- 238000000605 extraction Methods 0.000 claims description 6
- 230000004927 fusion Effects 0.000 claims description 6
- 239000011521 glass Substances 0.000 claims description 6
- 229920003229 poly(methyl methacrylate) Polymers 0.000 claims description 6
- 239000004926 polymethyl methacrylate Substances 0.000 claims description 6
- 238000012937 correction Methods 0.000 claims description 4
- 238000013507 mapping Methods 0.000 claims description 4
- 238000011084 recovery Methods 0.000 claims description 3
- 235000020183 skimmed milk Nutrition 0.000 claims description 3
- 238000010276 construction Methods 0.000 claims description 2
- 235000013336 milk Nutrition 0.000 claims description 2
- 239000008267 milk Substances 0.000 claims description 2
- 210000004080 milk Anatomy 0.000 claims description 2
- 230000003042 antagnostic effect Effects 0.000 claims 1
- 238000010586 diagram Methods 0.000 description 7
- 230000000694 effects Effects 0.000 description 5
- 238000005516 engineering process Methods 0.000 description 5
- 238000013135 deep learning Methods 0.000 description 3
- 230000004913 activation Effects 0.000 description 2
- 238000006731 degradation reaction Methods 0.000 description 2
- 238000011156 evaluation Methods 0.000 description 2
- 208000037170 Delayed Emergence from Anesthesia Diseases 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 230000015556 catabolic process Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000005286 illumination Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 238000012634 optical imaging Methods 0.000 description 1
- 239000002245 particle Substances 0.000 description 1
- 230000008569 process Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000016776 visual perception Effects 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/217—Validation; Performance evaluation; Active pattern learning techniques
-
- 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
- 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
- G06N3/048—Activation functions
-
- 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
-
- 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
-
- 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
- G06T2207/20084—Artificial neural networks [ANN]
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- General Health & Medical Sciences (AREA)
- Software Systems (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Mathematical Physics (AREA)
- Computational Linguistics (AREA)
- Health & Medical Sciences (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Image Processing (AREA)
Abstract
The invention provides a turbid underwater polarization image restoration method based on a generation countermeasure network. The method comprises the following steps: step 1, constructing an underwater active imaging system, and shooting a clear underwater intensity image and a polarized image under turbid water; step 2, establishing a data set; dividing a data set into a training set, a verification set and a test set according to the proportion of 8:1: 1; step 3, constructing a generating network; step 4, pre-training; pre-training the generated network by using the data set in the step two; step 5, constructing a discrimination network; step 6, constructing and generating a confrontation network; forming a generation countermeasure network with the generation network pre-trained in the fourth step, and training the generation countermeasure network by using the cross entropy as a loss function; and 7, restoring the image by using the trained generation countermeasure network. Compared with the prior art, the method can realize the restoration of the underwater polarization image with high turbidity.
Description
Technical Field
The invention relates to the technical field of deep learning and polarization imaging, in particular to a turbid underwater polarization image restoration method based on a generation countermeasure network.
Background
The underwater image is used as a main mode for researching an underwater world, and has an important role in research in various fields such as underwater energy exploration, underwater archaeology, underwater ecological environment monitoring protection, underwater military and the like. However, there are very many particles under water, and the imaging quality is seriously affected by the effect of the backward scattering light in the conventional underwater optical imaging, such as low contrast, low brightness, blurred object details, and the like. This makes image-based target detection and analysis difficult to use underwater. At present, underwater image restoration methods are mainly divided into two categories, including restoration methods based on physical models and image enhancement methods based on non-physical models. The restoration method based on the physical model needs to consider the degradation process of the image, perform mathematical modeling on the underwater imaging process, estimate model parameters to restore the image through inverse push, or utilize a deep learning technology to learn a mapping function to restore the image. The image enhancement method based on the non-physical model mainly comprises a histogram equalization method, a color correction method, a fusion-based method and the like. Both techniques improve the image to a certain extent and improve the image quality. In contrast, image enhancement generally enhances visual perception, and is biased towards subjective judgment of people, and lost detailed information cannot be repaired. The image restoration method is to perform image modeling according to the image degradation model and design a mapping function to restore the original image. The polarization imaging technology is an image restoration technology which is widely researched at present, but the imaging effect still has certain limitation, and further application is influenced.
Disclosure of Invention
In order to overcome the defects of the prior art, the invention provides a turbid underwater polarized image restoration method based on a generation countermeasure network by utilizing a polarized imaging technology and a deep learning technology. The network is used for restoring the image under the turbid water, so that the quality of the image can be effectively improved.
Technical scheme of the invention
A turbid underwater polarization image restoration method based on a generation countermeasure network comprises the following restoration steps:
step 4, pre-training; pre-training the generated network by using the data set in the step 2; the trained label is a clear intensity image of the underwater target corresponding to the polarization image; optimizing network parameters by using the mean square error as a loss function;
step 6, constructing and generating a confrontation network; forming a generation countermeasure network with the generation network pre-trained in the step four; respectively carrying out pair discrimination on the network and generating a network construction loss function by utilizing the cross entropy; the loss of the generated network is that the cross entropy of the generated picture is calculated by the output of the discrimination network and the label of the clear intensity image of the underwater target object; the loss of the discrimination network is composed of two parts, one part is the discrimination loss of the generated picture, the other part is the discrimination loss of the clear picture, and the two parts are added to obtain the total loss function of the discrimination network;
step 7, training the generation countermeasure network; and predicting the polarization image shot in the turbid water by the trained generated countermeasure network, and finally obtaining the obviously improved restored image.
Further, in the step 1, a clear intensity image of an underwater target is shot, light beams emitted by a light source sequentially pass through a first polarizing film, a first quarter-wave plate and a beam expander of a polarization modulation system and then irradiate on the underwater target, and the light beams are reflected by the target and then sequentially pass through a second polarizing film of an image acquisition system to reach the CMOS camera; thereby obtaining a clear intensity image of the underwater target.
Further, adding skim milk into water, controlling a second polarizer in front of the CMOS camera by using a stepping motor, and sequentially rotating to 0 degrees, 45 degrees and 90 degrees to obtain three polarization images with polarization directions of 0 degrees, 45 degrees and 90 degrees; and adding a second quarter-wave plate in front of the CMOS camera and the second polarizer to obtain a circular polarization image, thereby obtaining a polarization image of a target object in the turbid underwater environment.
Further, a 532nm blue-green laser is adopted as a light source.
Further, in step 3, the feature extraction is composed of 2 convolution layers, the size of each convolution kernel is 3, and the number of the convolution kernels is 64; the residual dense block includes 6 convolution layers densely connected with convolution kernels of size 3, 1 convolution layer with convolution kernel size 1, and a linear correction unit function (ReLU).
Further, in step 5, the convolution kernels of the 6 convolutional layers of the decision network are all 3 in size, the step size is 2, and the number of the convolution kernels is 64, 128, 256 and 256, respectively.
Further, the target is placed in a glass jar filled with water for simulating an underwater environment.
Furthermore, the glass cylinder is made of PMMA (polymethyl methacrylate).
The invention has the advantages that:
the method combines the generation countermeasure network, fully utilizes the polarization information of the image, pre-trains the generation network, trains the generation countermeasure network, and can effectively recover the image of the finally obtained generation countermeasure network, and the polarization image shot underwater with higher turbidity can still obtain good recovery effect by generating the countermeasure network.
Drawings
FIG. 1 is a schematic overall flow chart of the present invention.
Fig. 2 is a schematic diagram of the underwater active imaging system device of the invention.
Fig. 3 is a schematic diagram of a generating network structure according to the present invention.
FIG. 4 is a diagram of a dense residual block structure according to the present invention
Fig. 5 is a schematic diagram of the generation countermeasure network of the present invention.
Fig. 6 is a diagram of the experimental effect of the generation of the countermeasure network according to the present invention.
Reference numerals: the device comprises a light source 1, a first polaroid 2, a first 1/4 wave plate 3, a beam expander 4, a glass water jar 5, a simulated underwater environment 6, a target 7, a second 1/4 wave plate 8, a second polaroid 9 and a CMOS camera 10.
Detailed Description
In order to make the purpose and technical solution of the present invention clearer, the technical solution of the present invention is explained in detail below with reference to fig. 1 to 6.
The invention discloses a turbid underwater polarization image restoration method based on a generation countermeasure network, which comprises the following detailed steps:
As shown in fig. 2, a light beam emitted by a light source 1 sequentially passes through a first polarizer 2, a first quarter-wave plate 3 and a beam expander 4 of a polarization modulation system, then irradiates a target 7 in water, is reflected by the target 7, and then sequentially passes through a second polarizer 9 of an image acquisition system to reach a CMOS camera 10; shooting a clear intensity image of an underwater target; adding skim milk with different concentrations, controlling a second polaroid 9 in front of a CMOS camera 10 by using a stepping motor, and sequentially rotating to 0 degree, 45 degrees and 90 degrees to obtain three polarization images with polarization directions of 0 degree, 45 degrees and 90 degrees; a second quarter-wave plate 8 is added in front of the CMOS camera 10 and the second polarizer 9 to obtain a circularly polarized image. Specifically, in the present embodiment, 20 sets of images are captured, each set being added with 20 kinds of milk with different concentrations, and used for simulating the underwater environment 6 under 20 kinds of different turbidities, each turbidity being obtained by capturing 4 polarized images (including three polarized images with polarization directions of 0 °, 45 °, and 90 ° and one circularly polarized image) of the target 7 according to the capturing mode in step 1. A total of 12800(20 × 20 × 4 × 8) input images are obtained.
the feature extraction is composed of 2 convolution layers, the sizes of the convolution layers are all 3, and the number of the convolution layers is 64;
and constructing a residual error dense block, wherein the residual error dense block comprises a convolution layer with a convolution kernel of 3, a linear correction unit function (ReLU) and feature fusion. Fig. 4 is a schematic structural diagram of the residual dense block. In this embodiment, 4 residual dense blocks are used, each using 6 densely connected convolutional layers and corresponding activation functions: a linear modified unit function (ReLU);
and 4, pre-training the generated network by using the data set in the step 2, wherein the trained label is a clear intensity image of the underwater target corresponding to the polarization image, and the network parameters are optimized by using the mean square error (nn. MSELoss provided by the pytorch frame) as a loss function.
| Layer | Output Shape |
| Input | (batchsize,1,200,200) |
| Conv2d+LeakyReLU | (batchsize,64,100,100) |
| Conv2d+LeakyReLU | (batchsize,64,50,50) |
| Conv2d+LeakyReLU | (batchsize,128,25,25) |
| Conv2d+LeakyReLU | (batchsize,128,13,13) |
| Conv2d+LeakyReLU | (batchsize,256,7,7) |
| Conv2d+LeakyReLU | (batchsize,256,4,4) |
| Flatten | (batchsize,4096) |
| Linear | (batchsize,1) |
| Sigmoid | (batchsize,1) |
TABLE 1
Specifically, the discrimination network is composed of 6 convolutional layers, an activation function (LeakyReLu), a full connection layer and a sigmoid function, wherein the sizes of convolutional kernels of the 6 convolutional layers are all 3, the step lengths are all 2, and the number of convolutional kernels is 64, 128, 256 and 256 respectively. The input of the discrimination network is an image with the size of 200 x 200, the image size is compressed through 6 convolution layers, channels are amplified, and the image input into the discrimination network is judged to be a clear intensity image of an underwater target or an image restored by a generator.
And step 6, forming a generation countermeasure network with the generation network pre-trained in step 3, as shown in fig. 5. Cross entropy (nn. bceloss provided by the pytorech framework) was used to construct loss functions for the discriminative network and the generative network, respectively. The loss of the generated network is the cross entropy of the generated picture through the output of the discrimination network and the label calculation of the clear intensity image of the underwater target object. The loss function d _ loss of the discrimination network is: d _ loss _ real + d _ loss _ fake, wherein d _ loss _ real represents the loss of the clear intensity image of the underwater target object from the discrimination network, and d _ loss _ fake represents the loss of the image generated by the generation network from the discrimination network.
And 7, training the generated countermeasure network, predicting the polarization image shot in the turbid water by the trained generated countermeasure network, and finally obtaining the obviously improved restored image.
The experimental result shows that the method can effectively recover the polarization image shot in the high-turbidity water. The image restoration effect is obvious by combining objective evaluation criteria EME (the value of measure of evaluation), structural Similarity SSIM (structural Similarity Index measure) and subjective feeling. As shown in fig. 6, (a) is a high turbidity image, (b) is an image restored by the present invention, and (c) is a clear intensity image.
The EME and SSIM results are shown in table 2 below:
| high opacity images | The invention | Clear image | |
| EME | 1.02 | 14.54 | 16.85 |
| SSIM | 0.47 | 0.85 | 1 |
TABLE 2
It should be noted that although the embodiments of the present invention have been described in conjunction with the above-mentioned figures, the patentees can make variations and modifications within the scope of the appended claims, and within the scope of the present invention, as long as the protection scope of the present invention is not exceeded by the claims.
Claims (8)
1. A turbid underwater polarization image restoration method based on a generation countermeasure network is characterized by comprising the following restoration steps:
step 1, constructing an underwater active imaging system, and shooting a clear intensity image of an underwater target (7) and a polarization image of a turbid underwater target (7);
step 2, establishing a data set; cutting the polarization image obtained in the step 1, and dividing the polarization image into a training set, a verification set and a test set according to the ratio of 8:1:1 after a data set is expanded by overturning and rotating;
step 3, constructing a generating network; the generation network consists of feature extraction, residual error dense blocks, feature fusion and residual error learning; the method comprises the steps that a network is generated to have 4 inputs, namely a polarization image and a circular polarization image with the polarization directions of 0 degrees, 45 degrees and 90 degrees, 2 convolutional layers are used for feature extraction, 4 residual error dense blocks are used for extracting features with higher dimensionality, then feature fusion, residual error learning and one convolutional layer are used for carrying out nonlinear mapping on the 4 inputs, and finally a predicted recovery image is output;
step 4, pre-training; pre-training the generated network by using the data set in the step 2; the trained label is a clear intensity image of the underwater target corresponding to the polarization image; optimizing network parameters by using the mean square error as a loss function;
step 5, constructing a discrimination network; the input of the discrimination network is an image with the size of 200 x 200, the size of the image is compressed through 6 convolution layers, the channel is amplified, and the image input into the discrimination network is judged to be a clear intensity image of an underwater target or a predicted restored image output by the network generated in the third step;
step 6, constructing and generating a confrontation network; forming a generation countermeasure network with the generation network pre-trained in the step four; respectively carrying out pair discrimination on the network and generating a network construction loss function by utilizing the cross entropy; the loss of the generated network is that the cross entropy of the generated picture is calculated by the output of the discrimination network and the label of the clear intensity image of the underwater target object; the loss of the discrimination network is composed of two parts, one part is the discrimination loss of the generated picture, the other part is the discrimination loss of the clear picture, and the two parts are added to obtain the total loss function of the discrimination network;
step 7, training the generation countermeasure network; and predicting the polarization image shot in the turbid water by the trained generated countermeasure network, and finally obtaining the obviously improved restored image.
2. The method for restoring the underwater polarized image based on the turbidity generated by the countermeasure network according to claim 1, wherein in step 1, a clear intensity image of the underwater target is captured, and a light beam emitted by the light source (1) sequentially passes through the first polarizer (2), the first quarter-wave plate (3) and the beam expander (4) of the polarization modulation system, then irradiates on the underwater target (7), and is reflected by the target (7), and then sequentially passes through the second polarizer (9) of the image acquisition system and then reaches the CMOS camera (10), so as to obtain a clear intensity image of the underwater target.
3. The turbid underwater polarization image restoration method based on the antagonistic network generation is characterized in that skim milk is added into water, a second polarizer (9) in front of a CMOS camera (10) is controlled by a stepping motor, and the degreased milk is sequentially rotated to 0 degrees, 45 degrees and 90 degrees to obtain three polarization images with the polarization directions of 0 degrees, 45 degrees and 90 degrees; a second quarter-wave plate (8) is added in front of the CMOS camera (10) and the second polarizer (9) to obtain a circular polarization image; thereby obtaining a polarized image of the object (7) in the turbid underwater environment.
4. A turbid underwater polarization image restoration method based on generation of countermeasure networks according to one of the claims 2 to 3, characterized in that the light source (1) employs a 532nm blue-green laser.
5. The method for restoring the polarized image under the turbid water based on the generation countermeasure network, according to the claim 1, characterized in that in the step 3, the feature extraction is composed of 2 convolution layers, the convolution kernel size is 3, and the number of the convolution kernels is 64; the residual dense block includes 6 convolution layers densely connected with convolution kernels of size 3, 1 convolution layer with convolution kernel size 1, and a linear correction unit function (ReLU).
6. The method for restoring the polarized image under turbid water based on the generation countermeasure network of claim 1, wherein in step 5, the convolution kernels of 6 convolution layers of the discrimination network are all 3 in size and 2 in step size, and the number of the convolution kernels is 64, 128, 256 and 256 respectively.
7. A turbid underwater polarization image restoration method based on generation of countermeasure networks according to one of the claims 2 to 3, characterized in that the object (7) is placed in a glass jar (5) filled with water for simulating an underwater environment (6).
8. The turbid underwater polarization image restoration method based on generation of the countermeasure network according to claim 7, characterized in that the glass cylinder (5) is made of PMMA (polymethyl methacrylate).
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111514937.7A CN114170107A (en) | 2021-12-13 | 2021-12-13 | Turbid underwater polarization image restoration method based on generation countermeasure network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111514937.7A CN114170107A (en) | 2021-12-13 | 2021-12-13 | Turbid underwater polarization image restoration method based on generation countermeasure network |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114170107A true CN114170107A (en) | 2022-03-11 |
Family
ID=80485893
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111514937.7A Pending CN114170107A (en) | 2021-12-13 | 2021-12-13 | Turbid underwater polarization image restoration method based on generation countermeasure network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114170107A (en) |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20170351952A1 (en) * | 2016-06-01 | 2017-12-07 | Kla-Tencor Corporation | Systems and methods incorporating a neural network and a forward physical model for semiconductor applications |
| CN112950521A (en) * | 2021-04-27 | 2021-06-11 | 上海海事大学 | Image defogging method and generator network |
| CN113487504A (en) * | 2021-07-01 | 2021-10-08 | 天津大学 | Underwater gray level polarization image restoration method based on generation countermeasure network |
| CN113538279A (en) * | 2021-07-19 | 2021-10-22 | 浙江理工大学 | Underwater polarization image restoration system based on deep learning and restoration method thereof |
| US20210333151A1 (en) * | 2018-09-26 | 2021-10-28 | Nippon Telegraph And Telephone Corporation | Polarization Imaging System |
| CN113674190A (en) * | 2021-08-20 | 2021-11-19 | 中国人民解放军国防科技大学 | Image fusion method and device for generating countermeasure network based on dense connection |
-
2021
- 2021-12-13 CN CN202111514937.7A patent/CN114170107A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20170351952A1 (en) * | 2016-06-01 | 2017-12-07 | Kla-Tencor Corporation | Systems and methods incorporating a neural network and a forward physical model for semiconductor applications |
| US20210333151A1 (en) * | 2018-09-26 | 2021-10-28 | Nippon Telegraph And Telephone Corporation | Polarization Imaging System |
| CN112950521A (en) * | 2021-04-27 | 2021-06-11 | 上海海事大学 | Image defogging method and generator network |
| CN113487504A (en) * | 2021-07-01 | 2021-10-08 | 天津大学 | Underwater gray level polarization image restoration method based on generation countermeasure network |
| CN113538279A (en) * | 2021-07-19 | 2021-10-22 | 浙江理工大学 | Underwater polarization image restoration system based on deep learning and restoration method thereof |
| CN113674190A (en) * | 2021-08-20 | 2021-11-19 | 中国人民解放军国防科技大学 | Image fusion method and device for generating countermeasure network based on dense connection |
Non-Patent Citations (1)
| Title |
|---|
| 丛晓峰;章军;胡强;: "基于对偶学习的图像去雾网络", 应用光学, no. 01, 15 January 2020 (2020-01-15) * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108495110B (en) | Virtual viewpoint image generation method based on generation type countermeasure network | |
| CN110189278B (en) | Binocular scene image restoration method based on generation countermeasure network | |
| CN111292264B (en) | Image high dynamic range reconstruction method based on deep learning | |
| CN111062880A (en) | Underwater image real-time enhancement method based on condition generation countermeasure network | |
| CN109801232A (en) | A kind of single image to the fog method based on deep learning | |
| CN108573479A (en) | The facial image deblurring and restoration methods of confrontation type network are generated based on antithesis | |
| CN111612717A (en) | Water surface image reflection removing and repairing method and device based on countermeasure network | |
| CN108257101A (en) | A kind of underwater picture Enhancement Method based on optimal recovery parameter | |
| CN113160085B (en) | Water bloom shielding image data collection method based on generation countermeasure network | |
| Hou et al. | Reconstructing a high dynamic range image with a deeply unsupervised fusion model | |
| Zhang et al. | Enhanced visual perception for underwater images based on multistage generative adversarial network | |
| Mathai et al. | Transparent object reconstruction based on compressive sensing and super-resolution convolutional neural network | |
| CN116033279B (en) | Near infrared image colorization method, system and equipment for night monitoring camera | |
| CN115965844B (en) | Multi-focus image fusion method based on visual saliency priori knowledge | |
| CN114170107A (en) | Turbid underwater polarization image restoration method based on generation countermeasure network | |
| Krishnan et al. | A novel underwater image enhancement technique using ResNet | |
| CN114820395B (en) | Underwater image enhancement method based on multi-field information fusion | |
| CN113034432B (en) | Product defect detection method, system, device and storage medium | |
| CN114119428B (en) | Image deblurring method and device | |
| CN113538279A (en) | Underwater polarization image restoration system based on deep learning and restoration method thereof | |
| Jiang et al. | Haze relevant feature attention network for single image dehazing | |
| CN114758030A (en) | Underwater polarization imaging method integrating physical model and deep learning | |
| CN113160081A (en) | Depth face image restoration method based on perception deblurring | |
| Feng | A survey on video dehazing using deep learning | |
| Wang et al. | Real-World Atmospheric Turbulence Correction via Domain Adaptation |
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 |