CN110189268A - Underwater picture color correcting method based on GAN network - Google Patents
Underwater picture color correcting method based on GAN network Download PDFInfo
- Publication number
- CN110189268A CN110189268A CN201910432105.7A CN201910432105A CN110189268A CN 110189268 A CN110189268 A CN 110189268A CN 201910432105 A CN201910432105 A CN 201910432105A CN 110189268 A CN110189268 A CN 110189268A
- Authority
- CN
- China
- Prior art keywords
- image
- underwater
- underwater picture
- air
- network
- 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
- 238000000034 method Methods 0.000 title claims abstract description 27
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 claims abstract description 26
- 238000012549 training Methods 0.000 claims abstract description 15
- 238000012937 correction Methods 0.000 claims abstract description 8
- 239000000203 mixture Substances 0.000 claims abstract description 3
- 238000006243 chemical reaction Methods 0.000 claims description 7
- 239000000284 extract Substances 0.000 claims description 7
- 238000005457 optimization Methods 0.000 claims 1
- 238000012545 processing Methods 0.000 abstract description 4
- 238000013528 artificial neural network Methods 0.000 description 4
- 238000010586 diagram Methods 0.000 description 3
- 238000003384 imaging method Methods 0.000 description 3
- 230000000694 effects Effects 0.000 description 2
- 230000002708 enhancing effect Effects 0.000 description 2
- 238000004088 simulation Methods 0.000 description 2
- 238000012360 testing method Methods 0.000 description 2
- 241000251468 Actinopterygii Species 0.000 description 1
- 235000014653 Carica parviflora Nutrition 0.000 description 1
- 241000243321 Cnidaria Species 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000015556 catabolic process Effects 0.000 description 1
- 230000006835 compression Effects 0.000 description 1
- 238000007906 compression Methods 0.000 description 1
- 238000007796 conventional method Methods 0.000 description 1
- 238000013527 convolutional neural network Methods 0.000 description 1
- 238000013480 data collection Methods 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 230000004438 eyesight Effects 0.000 description 1
- 230000001902 propagating effect Effects 0.000 description 1
- 238000013139 quantization Methods 0.000 description 1
- 230000000717 retained effect Effects 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T3/00—Geometric image transformations in the plane of the image
- G06T3/04—Context-preserving transformations, e.g. by using an importance map
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/80—Geometric correction
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/90—Determination of colour characteristics
-
- 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
- G06T2207/10—Image acquisition modality
- G06T2207/10004—Still image; Photographic image
-
- 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]
-
- 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
- G06T2207/20212—Image combination
- G06T2207/20221—Image fusion; Image merging
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Image Processing (AREA)
Abstract
The present invention relates to a kind of underwater picture color correcting methods based on GAN network；This method passes through training Cycle GAN network, so that it is completed the image from air and converts to the style of underwater picture；Image in air is generated into corresponding underwater picture using trained Cycle GAN network, the data set of image underwater picture composition pairing corresponding with its in air；By constructing UGAN network, the underwater picture of the data set and actual photographed that use pairing collects training UGAN network as verifying, it obtains trained removing water model, using going water model to carry out colour correction to underwater picture, RGB image output is converted to after finally being merged the channel Y of image in the air of the channel UV of the underwater picture of input and output；The data set of the corresponding underwater picture pairing of image really can be used in present invention air generated, after carrying out water process to underwater picture, can enhance image quality, processing speed is fast, high-efficient.
Description
(1), technical field:
The present invention relates to a kind of image color correction method, in particular to a kind of underwater picture color based on GAN network
Bearing calibration.
(2), background technique:
Due to deep-sea shooting environmental complexity, imaging device captured image often will appear color offset, obscure, contrast
Low image quality issues reduce imaging device in ocean so that the visual experience of deep-sea image receives strong influence
Bio-identification, the computer visions such as target detection and tracking apply the performance of upper effect in water.
Image enhancement and image reconstruction both direction are broadly divided into the colour correction of underwater picture.Image enhancement direction
Algorithm mainly includes multi-scale image enhancing algorithm (the A Multiscale Retinex for Bridging based on Retinex
The Gap Between Color Images and the Human Observation of Scenes), this method exists
Under the hypothesis of Retinex theory by picture breakdown be irradiation component and reflecting component, on each channel to reflecting component carry out
The Gaussian Blur of different scale, biggish blur radius can be good at the detailed information for retaining image, lesser blur radius
The contrast information that image can be retained, finally again merges triple channel, and therefore, this method can improve the colour cast of underwater picture
Problem.But this method is more sensitive to noise spot when handling underwater picture, treated, and image will appear ring effect
It answers.The algorithm in image reconstruction direction include it is traditional based on dark channel prior and improved underwater picture algorithm for reconstructing and in recent years
Underwater picture algorithm for reconstructing neural network based.Paper Underwater Depth Estimation and Image
The underwater picture algorithm for reconstructing based on dark channel prior is proposed in Restoration Based on Single Images, is led to
The bias light and transmissivity in dark channel prior information solution atmospheric physics model are crossed, so that it is right to solve its according to underwater picture
The image in air answered, to achieve the purpose that water.However attenuation degree of light when propagating in water with propagate in air
When attenuation degree it is different, wavelength is longer to decay faster, therefore, underwater picture inclined blue-green mostly, so that dark channel prior
Information fails under water.Paper WaterGAN:Unsupervised Generative Network to Enable Real-
Water neural network based is proposed in time Color Correction of Monocular Underwater Images
Lower image reconstruction algorithm, because lacking true available paired data collection, this method generates underwater picture and its sky using GAN network
The data set that image matches in this way in gas carries out inverting reconstruction to underwater picture then using data set training CNN network.
But this method does not account for influence of the depth to underwater picture imaging system, manual simulation's ring when making data set
The underwater picture shot in border can not embody camera depth and obscure caused by image.
(3), summary of the invention:
The technical problem to be solved by the present invention is a kind of underwater picture color correcting method based on GAN network is provided, it should
Method produces the data set of the corresponding underwater picture pairing of image in air, and data set really can be used；To underwater picture
After carrying out water process, image quality can be enhanced, processing speed is fast, high-efficient.
Technical solution of the present invention:
A kind of underwater picture color correcting method based on GAN network, contains following steps:
Water model is removed in step 1, training network, foundation；
Step 1.1 prepares image in M underwater pictures and N air, forms unpaired data set, M is not less than 6500, N
Not less than 6000；
Step 1.2, building Cycle GAN network, the unpaired data set training Cycle GAN obtained using step 1.1
Network makes it complete the image from air and converts to the style of underwater picture；
Step 1.3 is generated image in the air in step 1.1 using Cycle GAN network trained in step 1.2
Its corresponding underwater picture, data set of the image underwater picture composition N corresponding with its to pairing in air；
Step 1.4 extracts P underwater pictures as verifying collection from the deep-sea video of actual photographed, and P is not less than 1000；
Step 1.5, building UGAN network；
Step 1.6 uses obtained N in step 1.3 to the data set of pairing as training set, gained in step 1.4
The P arrived underwater pictures are as verifying collection, constructed UGAN network in training step 1.5, obtain trained removing water model；
Step 2, using go water model correct underwater picture color；
K is opened underwater pictures to be processed as input picture by step 2.1, removes water mould using trained in step 1.6
Type carries out colour correction, and image in K air of output, K is the natural number more than or equal to 1；
K in step 2.1 underwater pictures to be processed are converted to yuv format from rgb format respectively by step 2.2, and
Extract the channel UV of input picture after format is converted；
Image in K obtained in step 2.1 air is converted to yuv format from rgb format respectively, and mentioned by step 2.3
The channel Y of image is exported after taking format to convert；
It is step 2.4, the channel UV of the obtained K of step 2.2 underwater picture and step 2.3 is obtained with K water
The channel Y of image is merged in the corresponding K air of lower image, obtains K YUV images；
K obtained in step 2.4 YUV images are converted into RGB image output by step 2.5.
In step 1.1, deep-sea video is downloaded from Youtube, and M underwater pictures are therefrom extracted using FFMPEG；
The image from the air for collecting N abyssopelagic organisms on ImageNet.
The loss function for the Cycle GAN network that step 1.2 is trained includes confrontation loss, MSE loss and circulation one
The loss of cause property, optimizer use Adm.
In step 1.4, FFMPEG is used to extract P underwater pictures from the deep-sea video of actual photographed as verifying
Collection.
The loss function for the UGAN network trained in step 1.6 includes confrontation loss and MSE loss.
In step 2.2 and step 2.3, image in underwater picture or air is converted into turning for yuv format from rgb format
Change formula are as follows:
Y=0.299R+0.587G+0.114
U=-0.1687R-0.3313G+0.5B+128
V=0.5R-0.4187G-0.0813B+128.
In step 2.5, image is converted to the conversion formula of rgb format from yuv format are as follows:
R=Y+1.402 (V-128)
G=Y-0.34414 (U-128) -0.71414 (V-128)
B=Y+1.772 (U-128).
Beneficial effects of the present invention:
1, the present invention carries out colour correction to underwater picture especially deep-sea image using UGAN network, by the underwater of input
The channel Y of image is merged in the channel UV of image and the air of output, and underwater picture enhances figure after carrying out water process
The contrast information of picture enriches the detailed information of image, and clarity with higher and acutance, can meet deep-sea video picture
The demand of matter enhancing, not only avoids the complexity using the conventional method based on atmospheric physics model when calculating transmissivity,
The existing unreliability based on neural network method manual simulation's data set is also avoided simultaneously.
2, the present invention generates the data of the corresponding underwater picture pairing of image in air using Cycle GAN network
Collection, and image processing speed is improved by compression quantization neural network, data set really can be used, and image processing efficiency is high.
(4), Detailed description of the invention:
Fig. 1 is the underwater picture color correcting method flow diagram based on GAN network；
Fig. 2 is the structural schematic diagram of Cycle GAN network；
Fig. 3 is the structural schematic diagram of UGAN network.
(5), specific embodiment:
Underwater picture color correcting method based on GAN network contains following steps (as shown in Figure 1):
Water model is removed in step 1, training network, foundation；
Step 1.1 prepares image in 6500 underwater pictures and 6000 air, forms unpaired data set；
Step 1.2, building Cycle GAN network (structure is as shown in Figure 2), the unpaired data obtained using step 1.1
Collect training Cycle GAN network, so that it is completed the image from air and converted to the style of underwater picture；
Step 1.3 is generated image in the air in step 1.1 using Cycle GAN network trained in step 1.2
Its corresponding underwater picture, image underwater picture corresponding with its forms the data sets of 6000 pairs of pairings in air；
The underwater picture of generation can be used as the input of Water Network, and image can be used as GroundTruth in air
(true picture correctly marked)；
Step 1.4 extracts 1000 underwater pictures as verifying collection from the deep-sea video of actual photographed；
Step 1.5, building UGAN network (structure is as shown in Figure 3)；
Step 1.6 uses the data sets of obtained 6000 pairs of pairings in step 1.3 as training set, institute in step 1.4
Obtained 1000 underwater pictures are as verifying collection, constructed UGAN network in training step 1.5, obtain trained removing water
Model；
Step 2, using go water model correct underwater picture color；
Step 2.1, using 500 underwater pictures to be processed as input picture, remove water using trained in step 1.6
Model carries out colour correction, exports image in 500 air；
500 underwater pictures to be processed in step 2.1 are converted to yuv format from rgb format respectively by step 2.2,
And extract the channel UV of input picture after format conversion；
Image in 500 air obtained in step 2.1 is converted to yuv format from rgb format respectively by step 2.3, and
Extract the channel Y that image is exported after format is converted；
It is step 2.4, the channel UV of obtained 500 underwater pictures of step 2.2 and step 2.3 is obtained with 500
The channel Y for opening image in corresponding 500 air of underwater picture is merged, and 500 YUV images are obtained；To enhance figure
The detailed information of picture.
500 YUV images obtained in step 2.4 are converted into RGB image output by step 2.5.
In step 1.1, deep-sea video is downloaded from Youtube, and 6500 underwater figures are therefrom extracted using FFMPEG
Picture, inclined blue-green, contrast are low and fuzzy with different degrees of details mostly for the underwater picture extracted from the video of deep-sea；From
Image in the air of 6000 abyssopelagic organisms, such as fish, coral are collected on ImageNet, the image is with color abundant and carefully
Information is saved, and there is good contrast.
The loss function for the Cycle GAN network that step 1.2 is trained includes confrontation loss, MSE (mean square error) loss
And circulation consistency loss, optimizer use Adm.
In step 1.4, FFMPEG is used to extract 1000 underwater pictures from the deep-sea video of actual photographed as testing
Card collection；
1000 are extracted from the deep-sea video really shot the degraded images such as colour cast, fuzzy and contrast be low occurs to test
Card goes the performance of Water Network.
The loss function for the UGAN network trained in step 1.6 includes confrontation loss and MSE (mean square error) loss.
In step 2.2 and step 2.3, image in underwater picture or air is converted into turning for yuv format from rgb format
Change formula are as follows:
Y=0.299R+0.587G+0.114
U=-0.1687R-0.3313G+0.5B+128
V=0.5R-0.4187G-0.0813B+128.
In step 2.5, image is converted to the conversion formula of rgb format from yuv format are as follows:
R=Y+1.402 (V-128)
G=Y-0.34414 (U-128) -0.71414 (V-128)
B=Y+1.772 (U-128).
Claims (7)
1. a kind of underwater picture color correcting method based on GAN network, it is characterized in that: containing following steps:
Water model is removed in step 1, training network, foundation；
Step 1.1 prepares image in M underwater pictures and N air, forms unpaired data set, M is not small not less than 6500, N
In 6000；
Step 1.2, building Cycle GAN network, the unpaired data set training Cycle GAN network obtained using step 1.1,
Make it complete the image from air to convert to the style of underwater picture；
Step 1.3, that image in the air in step 1.1 generated its using Cycle GAN network trained in step 1.2 is right
The underwater picture answered, data set of the image underwater picture composition N corresponding with its to pairing in air；
Step 1.4 extracts P underwater pictures as verifying collection from the deep-sea video of actual photographed, and P is not less than 1000；
Step 1.5, building UGAN network；
Step 1.6 uses obtained N in step 1.3 to the data set of pairing as training set, obtained P in step 1.4
Underwater picture is as verifying collection, constructed UGAN network in training step 1.5, obtains trained removing water model；
Step 2, using go water model correct underwater picture color；
Step 2.1, using underwater pictures K to be processed as input picture, using in step 1.6 it is trained go water model into
Row colour correction, output K open image in air, and K is the natural number more than or equal to 1；
K in step 2.1 underwater pictures to be processed are converted to yuv format from rgb format respectively, and extracted by step 2.2
The channel UV of input picture after format conversion；
Image in K obtained in step 2.1 air is converted to yuv format from rgb format respectively, and extracts lattice by step 2.3
The channel Y of image is exported after formula conversion；
Step 2.4, the channel UV that the obtained K of step 2.2 is opened to underwater picture and obtained open with K of step 2.3 are schemed under water
As the channel Y of image in corresponding K air is merged, K YUV images are obtained；
K obtained in step 2.4 YUV images are converted into RGB image output by step 2.5.
2. the underwater picture color correcting method according to claim 1 based on GAN network, it is characterized in that: in the step
In rapid 1.1, deep-sea video is downloaded from Youtube, and M underwater pictures are therefrom extracted using FFMPEG；From ImageNet
Collect image in the air of N abyssopelagic organisms.
3. the underwater picture color correcting method according to claim 1 based on GAN network, it is characterized in that: the step
The loss function of the 1.2 Cycle GAN networks trained includes confrontation loss, MSE loss and circulation consistency loss, optimization
Device uses Adm.
4. the underwater picture color correcting method according to claim 1 based on GAN network, it is characterized in that: in the step
In rapid 1.4, uses FFMPEG to extract P underwater pictures from the deep-sea video of actual photographed and collect as verifying.
5. the underwater picture color correcting method according to claim 1 based on GAN network, it is characterized in that: the step
The loss function for the UGAN network trained in 1.6 includes confrontation loss and MSE loss.
6. the underwater picture color correcting method according to claim 1 based on GAN network, it is characterized in that: in the step
Rapid 2.2 and step 2.3 in, image in underwater picture or air is converted to the conversion formula of yuv format from rgb format are as follows:
Y=0.299 R+0.587 G+0.114
U=-0.1687 R-0.3313 G+0.5 B+128
V=0.5 R-0.4187 G-0.0813 B+128.
7. the underwater picture color correcting method according to claim 1 based on GAN network, it is characterized in that: in the step
In rapid 2.5, image is converted to the conversion formula of rgb format from yuv format are as follows:
R=Y+1.402 (V-128)
G=Y-0.34414 (U-128) -0.71414 (V-128)
B=Y+1.772 (U-128).
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910432105.7A CN110189268A (en) | 2019-05-23 | 2019-05-23 | Underwater picture color correcting method based on GAN network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910432105.7A CN110189268A (en) | 2019-05-23 | 2019-05-23 | Underwater picture color correcting method based on GAN network |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110189268A true CN110189268A (en) | 2019-08-30 |
Family
ID=67717542
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910432105.7A Pending CN110189268A (en) | 2019-05-23 | 2019-05-23 | Underwater picture color correcting method based on GAN network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110189268A (en) |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111275647A (en) * | 2020-01-21 | 2020-06-12 | 南京信息工程大学 | Underwater image restoration method based on cyclic generation countermeasure network |
| CN111784753A (en) * | 2020-07-03 | 2020-10-16 | 江苏科技大学 | Three-dimensional reconstruction stereo matching method for autonomous underwater robot recovery butt joint foreground view field |
| CN112488130A (en) * | 2020-12-17 | 2021-03-12 | 苏州聚悦信息科技有限公司 | AI micro-pore wall detection algorithm |
| WO2021159633A1 (en) * | 2020-02-13 | 2021-08-19 | 平安科技（深圳）有限公司 | Method and system for training image recognition model, and image recognition method |
| CN113537401A (en) * | 2021-08-13 | 2021-10-22 | 哈尔滨工程大学 | Underwater image translation method based on improved generation countermeasure network |
| CN114429431A (en) * | 2022-04-01 | 2022-05-03 | 西南科技大学 | Identification method and system for converting image from underwater to air |
| CN114513684A (en) * | 2020-11-16 | 2022-05-17 | 飞狐信息技术（天津）有限公司 | Method for constructing video image quality enhancement model, method and device for enhancing video image quality |
| CN114528920A (en) * | 2022-01-19 | 2022-05-24 | 西北大学 | Terracotta soldiers and horses color restoration method based on generation of confrontation network |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108038832A (en) * | 2017-12-25 | 2018-05-15 | 中国科学院深圳先进技术研究院 | A kind of underwater picture Enhancement Method and system |
| CN109064422A (en) * | 2018-07-17 | 2018-12-21 | 中国海洋大学 | A kind of underwater image restoration method based on fusion confrontation network |
| CN109671124A (en) * | 2018-12-27 | 2019-04-23 | 湖南师范大学 | A kind of mineral flotation foam image color calibration method and foam color detection system |
| CN109767399A (en) * | 2019-01-14 | 2019-05-17 | 哈尔滨工程大学 | A kind of underwater picture Enhancement Method based on unsupervised colour correction |
-
2019
- 2019-05-23 CN CN201910432105.7A patent/CN110189268A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108038832A (en) * | 2017-12-25 | 2018-05-15 | 中国科学院深圳先进技术研究院 | A kind of underwater picture Enhancement Method and system |
| CN109064422A (en) * | 2018-07-17 | 2018-12-21 | 中国海洋大学 | A kind of underwater image restoration method based on fusion confrontation network |
| CN109671124A (en) * | 2018-12-27 | 2019-04-23 | 湖南师范大学 | A kind of mineral flotation foam image color calibration method and foam color detection system |
| CN109767399A (en) * | 2019-01-14 | 2019-05-17 | 哈尔滨工程大学 | A kind of underwater picture Enhancement Method based on unsupervised colour correction |
Non-Patent Citations (4)
| Title |
|---|
| CAMERON FABBRI ET AL: "Enhancing Underwater Imagery Using Generative Adversarial Networks", 《2018 IEEE INTERNATIONAL CONFERENCE ON ROBOTICS AND AUTOMATION (ICRA)》 * |
| YAN ZUO ET AL.: "Generative Adversarial Forests for Better Conditioned Adversarial Learning", 《MACHINE LEARNING》 * |
| 张晶: "《多媒体信息与通信》", 31 December 2017 * |
| 李庆忠: "基于小波变换的水下降质图像复原算法", 《计算机工程》 * |
Cited By (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111275647A (en) * | 2020-01-21 | 2020-06-12 | 南京信息工程大学 | Underwater image restoration method based on cyclic generation countermeasure network |
| WO2021159633A1 (en) * | 2020-02-13 | 2021-08-19 | 平安科技（深圳）有限公司 | Method and system for training image recognition model, and image recognition method |
| CN111784753A (en) * | 2020-07-03 | 2020-10-16 | 江苏科技大学 | Three-dimensional reconstruction stereo matching method for autonomous underwater robot recovery butt joint foreground view field |
| CN111784753B (en) * | 2020-07-03 | 2023-12-05 | 江苏科技大学 | Jing Shichang three-dimensional reconstruction stereo matching method before recovery and docking of autonomous underwater robot |
| CN114513684A (en) * | 2020-11-16 | 2022-05-17 | 飞狐信息技术（天津）有限公司 | Method for constructing video image quality enhancement model, method and device for enhancing video image quality |
| CN114513684B (en) * | 2020-11-16 | 2024-05-28 | 飞狐信息技术（天津）有限公司 | Method for constructing video image quality enhancement model, video image quality enhancement method and device |
| CN112488130B (en) * | 2020-12-17 | 2023-08-15 | 苏州聚悦信息科技有限公司 | AI micro hole wall detection method |
| CN112488130A (en) * | 2020-12-17 | 2021-03-12 | 苏州聚悦信息科技有限公司 | AI micro-pore wall detection algorithm |
| CN113537401A (en) * | 2021-08-13 | 2021-10-22 | 哈尔滨工程大学 | Underwater image translation method based on improved generation countermeasure network |
| CN113537401B (en) * | 2021-08-13 | 2022-08-02 | 哈尔滨工程大学 | Underwater image translation method based on improved generation countermeasure network |
| CN114528920A (en) * | 2022-01-19 | 2022-05-24 | 西北大学 | Terracotta soldiers and horses color restoration method based on generation of confrontation network |
| CN114528920B (en) * | 2022-01-19 | 2023-08-18 | 西北大学 | Terra-cotta warriors color restoration method based on generation countermeasure network |
| CN114429431A (en) * | 2022-04-01 | 2022-05-03 | 西南科技大学 | Identification method and system for converting image from underwater to air |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110189268A (en) | Underwater picture color correcting method based on GAN network | |
| CN110428366A (en) | Image processing method and device, electronic equipment, computer readable storage medium | |
| CN110473185A (en) | Image processing method and device, electronic equipment, computer readable storage medium | |
| Wang et al. | Joint iterative color correction and dehazing for underwater image enhancement | |
| CN110443763A (en) | A kind of Image shadow removal method based on convolutional neural networks | |
| CN111709900A (en) | High dynamic range image reconstruction method based on global feature guidance | |
| CN110225260B (en) | Three-dimensional high dynamic range imaging method based on generation countermeasure network | |
| CN115393227B (en) | Low-light full-color video image self-adaptive enhancement method and system based on deep learning | |
| Moriwaki et al. | Hybrid loss for learning single-image-based HDR reconstruction | |
| CN111724317A (en) | Method for constructing Raw domain video denoising supervision data set | |
| CN113284061B (en) | Underwater image enhancement method based on gradient network | |
| WO2023066173A1 (en) | Image processing method and apparatus, and storage medium and electronic device | |
| CN116563183A (en) | High dynamic range image reconstruction method and system based on single RAW image | |
| CN112508812A (en) | Image color cast correction method, model training method, device and equipment | |
| CN115035010A (en) | Underwater image enhancement method based on convolutional network guided model mapping | |
| CN111553856A (en) | Image defogging method based on depth estimation assistance | |
| CN112561782B (en) | Method for improving reality degree of simulation picture of offshore scene | |
| CN113554568A (en) | Unsupervised circulating rain removal network method based on self-supervision constraint and unpaired data | |
| CN107454318A (en) | Image processing method, device, mobile terminal and computer-readable recording medium | |
| CN112489103A (en) | High-resolution depth map acquisition method and system | |
| CN107392870A (en) | Image processing method, device, mobile terminal and computer-readable recording medium | |
| CN111161189A (en) | Single image re-enhancement method based on detail compensation network | |
| CN116563133A (en) | Low-illumination color image enhancement method based on simulated exposure and multi-scale fusion | |
| Zhang et al. | Single image dehazing via reinforcement learning | |
| CN116228550A (en) | Image self-enhancement defogging algorithm based on generation of countermeasure 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 | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190830 |