CN108717698A - A kind of high quality graphic generation method generating confrontation network based on depth convolution - Google Patents
A kind of high quality graphic generation method generating confrontation network based on depth convolution Download PDFInfo
- Publication number
- CN108717698A CN108717698A CN201810520114.7A CN201810520114A CN108717698A CN 108717698 A CN108717698 A CN 108717698A CN 201810520114 A CN201810520114 A CN 201810520114A CN 108717698 A CN108717698 A CN 108717698A
- Authority
- CN
- China
- Prior art keywords
- spot
- image
- loss function
- drop
- sar
- 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.)
- Withdrawn
Links
- 238000000034 method Methods 0.000 title claims abstract description 54
- 230000008485 antagonism Effects 0.000 claims abstract description 12
- 238000011156 evaluation Methods 0.000 claims abstract description 10
- 230000000007 visual effect Effects 0.000 claims abstract description 9
- 230000015572 biosynthetic process Effects 0.000 claims abstract description 6
- 238000003786 synthesis reaction Methods 0.000 claims abstract description 6
- 230000008569 process Effects 0.000 claims abstract description 4
- 230000011218 segmentation Effects 0.000 claims description 12
- 230000001427 coherent effect Effects 0.000 claims description 4
- 238000005516 engineering process Methods 0.000 claims description 4
- 238000012360 testing method Methods 0.000 claims description 4
- 208000003351 Melanosis Diseases 0.000 claims description 3
- 238000005562 fading Methods 0.000 claims description 3
- 238000003709 image segmentation Methods 0.000 claims description 3
- 238000013507 mapping Methods 0.000 claims description 3
- 238000012549 training Methods 0.000 claims description 3
- 238000013527 convolutional neural network Methods 0.000 claims description 2
- 238000003475 lamination Methods 0.000 claims 1
- 238000011056 performance test Methods 0.000 claims 1
- 238000012545 processing Methods 0.000 description 4
- 238000001514 detection method Methods 0.000 description 3
- 238000004458 analytical method Methods 0.000 description 2
- 238000003384 imaging method Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000012544 monitoring process Methods 0.000 description 2
- 238000003672 processing method Methods 0.000 description 2
- 238000009825 accumulation Methods 0.000 description 1
- 230000001154 acute effect Effects 0.000 description 1
- 230000008901 benefit Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000009545 invasion Effects 0.000 description 1
- 238000011835 investigation Methods 0.000 description 1
- 238000005065 mining Methods 0.000 description 1
- 230000001537 neural effect Effects 0.000 description 1
- 230000008520 organization Effects 0.000 description 1
- 230000002265 prevention Effects 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 239000013535 sea water Substances 0.000 description 1
- 239000002689 soil Substances 0.000 description 1
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/11—Region-based segmentation
-
- 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
- G06N3/08—Learning methods
-
- 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
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10032—Satellite or aerial image; Remote sensing
- G06T2207/10044—Radar 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
- G06T2207/20212—Image combination
- G06T2207/20221—Image fusion; Image merging
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Radar Systems Or Details Thereof (AREA)
Abstract
A kind of high quality graphic generation method generating confrontation network based on depth convolution proposed in the present invention, its main contents includes drop spot network, coloured networks, loss function and Performance Evaluation index, its process is, first using drop spot network to synthetic aperture radar (SAR) SAR) noise image that observes carries out drop spot and handles, then visual picture is converted by spot network is dropped with coloured networks；Next, (combining L by loss function1The characteristic of loss function and antagonism loss function) local grey is identified and is coloured, Performance Evaluation finally is carried out to this method using three drop spot performance, composograph result and true SAR image result indexs.The image artifacts that this method has compared existing method synthesis are less, protect detailed information；The appearance of local grey is also avoided in final output, can generate high quality visual picture.
Description
Technical field
The present invention relates to image processing fields, and the high-quality of confrontation network is generated based on depth convolution more particularly, to a kind of
Measure image generating method.
Background technology
Synthetic aperture radar (SAR) can by emit microwave, receive ground target reflection echo come obtain information carry out at
As (SAR image), this imaging technique is influenced small by weather and time, thus is widely used.But due to imaging radar
Transmitting is pure coherent wave, and when this signal irradiates target, the random scatter signal of target and the interference of transmitting signal generate spot
Spot noise, and make the grey scale pixel value acute variation of image, i.e., in uniform target surface, some pixels are presented bright spot, have
In dim spot, the fine structure of image has been obscured, has made the reduction of image interpretation ability, it is therefore, right in order to generate the image of high quality
SAR image handle particularly important.In agricultural field, SAR image can be applied to land use condition survey, crop and trees
All various aspects such as classification, crop and tree vigor detection, crop yield and the estimation of trees accumulation bring considerable economic effect
Benefit；In hydrology field, SAR image can be used for monitoring soil moisture, drainage analysis, Investigation of water resources, seawater invasion monitoring, river
Analysis on Vicissitudes etc., and achieve many important achievements in research；In addition, SAR image also the identification of military target with ruin
Hinder measures of effectiveness, the detection of mining deposits, the condition of a disaster detection and extensive use in the fields such as prevention, Medical Image Processing.However, existing
Pseudomorphism is more existing for the handling result of some SAR image processing methods, has accidentally deleted many picture detail information and has had part
Grey generates, and whole drop spot performance is undesirable.
The present invention proposes a kind of high quality graphic generation method generating confrontation network based on depth convolution, first using drop
Spot network is to synthetic aperture radar (SAR) SAR) noise image that observes carries out drop spot processing, and then spot net will drop with coloured networks
Network is converted into visual picture；Next, (combining L by loss function1The characteristic of loss function and antagonism loss function)
Local grey is identified and is coloured, three drop spot performance, composograph result and true SAR image result fingers are finally utilized
Mark carries out Performance Evaluation to this method.The image artifacts that this method has compared existing method synthesis are less, protect detailed information；
The appearance of local grey is also avoided in final output, can generate high quality visual picture.
Invention content
Pseudomorphism is more existing for handling result for existing SAR image processing method, has accidentally deleted many picture details
Information and there is local grey to generate, whole to drop the problems such as spot performance is undesirable, the purpose of the present invention is to provide one kind to be based on
Depth convolution generates the high quality graphic generation method of confrontation network, first using drop spot network to synthetic aperture radar (SAR) SAR) it sees
The noise image measured carries out drop spot processing, is then converted into visual picture by spot network is dropped with coloured networks；Next, passing through
Loss function (combines L1The characteristic of loss function and antagonism loss function) local grey is identified and is coloured, finally
Performance Evaluation is carried out to this method using three drop spot performance, composograph result and true SAR image result indexs.
To solve the above problems, the present invention provides a kind of high quality graphic generation generating confrontation network based on depth convolution
Method, main contents include：
(1) spot network drops；
(2) coloured networks；
(3) loss function；
(4) Performance Evaluation index.
Wherein, the drop spot network, it is main that noise synthetic aperture radar (SAR) image is handled using segmentation residual error method
(figure spot product)；Segmentation residual error method be by will the segmentation residual error layer of recognition element be incorporated in convolutional network, such convolutional layer
It can learn spot element in the training process；Spot is estimated in output representative before dividing residual error layer, and drop spot image is to pass through
Estimating for input is obtained after spot image segmentation.
Further, the SAR image, SAR are a kind of coherent radar image technologies, can be generated high-resolution
Image；SAR image can be indicated by following model：
Wherein, Y represents SAR image；F is standard colour fading spot noise stochastic variable；Refer to array element to be multiplied.
Further, the noise prediction segment generation estimated spot, be by drop spot network；Noise prediction segment is by 8
A convolutional layer composition, these convolutional layers have appropriate zero padding to ensure the image output and input dimension having the same；
First, changed to reduce internal covariance with batch standardization；Every layer of convolutional layer (in addition to the last one convolutional layer) includes 64 mistakes
Filter；Then, the segmentation residual error layer with jump connection is divided input picture with spot is estimated；Finally, a hyperbolic tangent
Layer is deposited in network end-point, as a non-linear formula.
Wherein, the coloured networks constitute symmetrical encoder-decoder with 8 convolutional layers and 3 jump connections
Convolutional neural networks；Each convolutional layer kernel size is 3 × 3；There are 64 Feature Mappings in each of the steps.
Wherein, the loss function, loss function here is by L one by one1Loss function and antagonism lose letter
Number is formed with appropriate weight, and function formula is defined as follows：
LD=LL1(gray(Y),gray(X)；GD) (2)
L=LD+LC (4)
Wherein, gray (X) and gray (Y) is respectively the gray scale form of X and Y；LDAnd LCSpot network and colored net respectively drop
Network；λaTo balance L1The weight of loss function and antagonism loss function.
Wherein, the Performance Evaluation index, in order to assess the efficiency and performance of this method, to this method at following three
It tests and assesses under index：Spot performance, composograph result and true SAR image result drop.
Further, the drop spot performance, mainly passes through Y-PSNR, structure similarity index, universal qualities index
The drop spot performance of this method is evaluated and tested with drop four indexs of spot gain.
Further, the composograph as a result, this method synthesis image pseudomorphism it is few, also done during nti-freckle
The protection to detailed information is arrived；The appearance that local grey is also avoided in final output, mainly due to this method
Combine L1The characteristic of loss function and antagonism loss function.
Further, the true SAR image result, which is characterized in that scheme the SAR of same place same time
As treated that image is compared for, satellite image and this method, as a result show that this method can be from SAR image generation high quality
Visual picture.
Description of the drawings
Fig. 1 is a kind of system framework for the high quality graphic generation method generating confrontation network based on depth convolution of the present invention
Figure.
Fig. 2 is a kind of network architecture for the high quality graphic generation method generating confrontation network based on depth convolution of the present invention
Figure.
Fig. 3 is a kind of handling result for the high quality graphic generation method generating confrontation network based on depth convolution of the present invention
Figure.
Specific implementation mode
It should be noted that in the absence of conflict, the features in the embodiments and the embodiments of the present application can phase
It mutually combines, invention is further described in detail in the following with reference to the drawings and specific embodiments.
Fig. 1 is a kind of system framework for the high quality graphic generation method generating confrontation network based on depth convolution of the present invention
Figure.Main includes drop spot network, coloured networks, loss function and Performance Evaluation index.
Loss function, loss function here is by L one by one1Loss function and antagonism loss function are with appropriate power
It reassembles into, function formula is defined as follows：
LD=LL1(gray(Y),gray(X)；GD) (1)
L=LD+LC (3)
Wherein, gray (X) and gray (Y) is respectively the gray scale form of X and Y；LDAnd LCSpot network and colored net respectively drop
Network；λaTo balance L1The weight of loss function and antagonism loss function.
Performance Evaluation index carries out this method to assess the efficiency and performance of this method under following three indexs
Test and appraisal：Spot performance, composograph result and true SAR image result drop.
Wherein, spot performance is dropped, Y-PSNR, structure similarity index, universal qualities index and drop spot gain are mainly passed through
Four indexs evaluate and test the drop spot performance of this method.
Wherein, composograph as a result, this method synthesis image pseudomorphism it is few, also accomplished to details during nti-freckle
The protection of information；The appearance of local grey is also avoided in final output, mainly since this method combines L1Damage
Lose the characteristic of function and antagonism loss function.
Wherein, true SAR image result, which is characterized in that by the SAR image of same place same time, satellite image
Treated that image is compared with this method, as a result shows that this method can generate high quality visual picture from SAR image.
Fig. 2 is a kind of network architecture for the high quality graphic generation method generating confrontation network based on depth convolution of the present invention
Figure.Organization Chart including dropping spot network and coloured networks.
Spot network drops, and it is main that noise synthetic aperture radar (SAR) image (figure spot product) is handled using segmentation residual error method；
Segmentation residual error method be by will the segmentation residual error layer of recognition element be incorporated in convolutional network, such convolutional layer can be in training
Learn spot element in the process；Spot is estimated in output representative before dividing residual error layer, and drop spot image is by being estimated to input
It is obtained after spot image segmentation.
Wherein, SAR image, SAR are a kind of coherent radar image technologies, can generate high-resolution image；SAR schemes
As that can be indicated by following model：
Wherein, Y represents SAR image；F is standard colour fading spot noise stochastic variable；Refer to array element to be multiplied.
Wherein, estimate spot, which is characterized in that generated by the noise prediction segment of drop spot network；Noise prediction segment is by 8
A convolutional layer composition, these convolutional layers have appropriate zero padding to ensure the image output and input dimension having the same；
First, changed to reduce internal covariance with batch standardization；Every layer of convolutional layer (in addition to the last one convolutional layer) includes 64 mistakes
Filter；Then, the segmentation residual error layer with jump connection is divided input picture with spot is estimated；Finally, a hyperbolic tangent
Layer is deposited in network end-point, as a non-linear formula.
Coloured networks constitute symmetrical encoder-decoder convolutional Neural net with 8 convolutional layers and 3 jump connections
Network；Each convolutional layer kernel size is 3 × 3；There are 64 Feature Mappings in each of the steps.
Fig. 3 is a kind of handling result for the high quality graphic generation method generating confrontation network based on depth convolution of the present invention
Figure.The image artifacts that existing method synthesis has been compared this figure shows this method are less, protect detailed information；In final output
As a result the appearance of local grey is also avoided in, can generate high quality visual picture.
For those skilled in the art, the present invention is not limited to the details of above-described embodiment, in the essence without departing substantially from the present invention
In the case of refreshing and range, the present invention can be realized in other specific forms.In addition, those skilled in the art can be to this hair
Bright to carry out various modification and variations without departing from the spirit and scope of the present invention, these improvements and modifications also should be regarded as the present invention's
Protection domain.Therefore, the following claims are intended to be interpreted as including preferred embodiment and falls into all changes of the scope of the invention
More and change.
Claims (10)
1. a kind of high quality graphic generation method generating confrontation network based on depth convolution, which is characterized in that main includes drop
Spot network (one)；Coloured networks (two)；Loss function (three)；Performance Evaluation index (four).
2. based on the drop spot network (one) described in claims 1, which is characterized in that main to be made an uproar using segmentation residual error method to handle
Phonosynthesis aperture radar (SAR) image (figure spot product)；Segmentation residual error method be by will recognition element segmentation residual error layer simultaneously
Enter in convolutional network, such convolutional layer can learn spot element in the training process；Output generation before dividing residual error layer
Table estimates spot, drop spot image be by input estimate spot image segmentation after obtain.
3. based on the SAR image described in claims 3, which is characterized in that SAR is a kind of coherent radar image technology, energy
Enough generate high-resolution image；SAR image can be indicated by following model：
Y=F ⊙ X (1)
Wherein, Y represents SAR image；F is standard colour fading spot noise stochastic variable；⊙ refers to array element multiplication.
4. based on spot is estimated described in claims 3, which is characterized in that generated by the noise prediction segment of drop spot network；
Noise prediction segment is made of 8 convolutional layers, these convolutional layers have appropriate zero padding to ensure the image output and input
Dimension having the same；First, changed to reduce internal covariance with batch standardization；Every layer of convolutional layer is (in addition to the last one volume
Lamination) include 64 filters；Then, the segmentation residual error layer with jump connection is divided input picture with spot is estimated；Finally,
One hyperbolic tangent layer is deposited in network end-point, as a non-linear formula.
5. based on the coloured networks (two) described in claims 1, which is characterized in that 8 convolutional layers of coloured networks and 3 jumps
Jump connection constitutes symmetrical encoder-decoder convolutional neural networks；Each convolutional layer kernel size is 3 × 3；Every
There are 64 Feature Mappings in one step.
6. based on the loss function (three) described in claims 1, which is characterized in that loss function here is by L one by one1
Loss function and antagonism loss function are formed with appropriate weight, and function formula is defined as follows：
LD=LL1(gray (Y), gray (X)；GD) (2)
L=LD+LC (4)
Wherein, gray (X) and gray (Y) is respectively the gray scale form of X and Y；LDAnd LCSpot network and coloured networks respectively drop；
λaTo balance L1The weight of loss function and antagonism loss function.
7. based on the Performance Evaluation index (four) described in claims 1, which is characterized in that in order to assess this method efficiency and
Performance tests and assesses under following three indexs to this method：Spot performance, composograph result and true SAR image result drop.
8. based on the drop spot performance described in claims 7, which is characterized in that mainly pass through Y-PSNR, the similar finger of structure
Number, universal qualities index and drop four indexs of spot gain evaluate and test the drop spot performance of this method.
9. based on the composograph result described in claims 7, which is characterized in that the pseudomorphism of the image of this method synthesis is few,
The protection to detailed information is also accomplished during nti-freckle；The appearance of local grey is also avoided in final output,
Mainly since this method combines L1The characteristic of loss function and antagonism loss function.
10. based on the true SAR image result described in claims 7, which is characterized in that by the same place same time
Treated that image is compared for SAR image, satellite image and this method, as a result shows that this method can be generated from SAR image
High quality visual picture.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810520114.7A CN108717698A (en) | 2018-05-28 | 2018-05-28 | A kind of high quality graphic generation method generating confrontation network based on depth convolution |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810520114.7A CN108717698A (en) | 2018-05-28 | 2018-05-28 | A kind of high quality graphic generation method generating confrontation network based on depth convolution |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108717698A true CN108717698A (en) | 2018-10-30 |
Family
ID=63900536
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810520114.7A Withdrawn CN108717698A (en) | 2018-05-28 | 2018-05-28 | A kind of high quality graphic generation method generating confrontation network based on depth convolution |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108717698A (en) |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109636742A (en) * | 2018-11-23 | 2019-04-16 | 中国人民解放军空军研究院航空兵研究所 | The SAR image of network and the mode conversion method of visible images are generated based on confrontation |
| CN109658344A (en) * | 2018-11-12 | 2019-04-19 | 哈尔滨工业大学（深圳） | Image de-noising method, device, equipment and storage medium based on deep learning |
| CN109859147A (en) * | 2019-03-01 | 2019-06-07 | 武汉大学 | A kind of true picture denoising method based on generation confrontation network noise modeling |
| CN110163801A (en) * | 2019-05-17 | 2019-08-23 | 深圳先进技术研究院 | A kind of Image Super-resolution and color method, system and electronic equipment |
| CN110751630A (en) * | 2019-09-30 | 2020-02-04 | 山东信通电子股份有限公司 | Power transmission line foreign matter detection method and device based on deep learning and medium |
| 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 |
| CN111862253A (en) * | 2020-07-14 | 2020-10-30 | 华中师范大学 | Sketch coloring method and system for generating confrontation network based on deep convolution |
| CN115546351A (en) * | 2022-12-02 | 2022-12-30 | 耕宇牧星(北京)空间科技有限公司 | Convolution network-based synthetic aperture radar image coloring method |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102368332A (en) * | 2011-11-17 | 2012-03-07 | 西安电子科技大学 | SAR (Synthetic Aperture Radar) image despeckling method based on local Gaussian model in NSCT (Non-subsampled Contourlet Transform) region |
| CN103761709A (en) * | 2014-01-07 | 2014-04-30 | 西安电子科技大学 | Parallel real-time SAR image spot and noise reducing method based on multiple DSPs |
-
2018
- 2018-05-28 CN CN201810520114.7A patent/CN108717698A/en not_active Withdrawn
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102368332A (en) * | 2011-11-17 | 2012-03-07 | 西安电子科技大学 | SAR (Synthetic Aperture Radar) image despeckling method based on local Gaussian model in NSCT (Non-subsampled Contourlet Transform) region |
| CN103761709A (en) * | 2014-01-07 | 2014-04-30 | 西安电子科技大学 | Parallel real-time SAR image spot and noise reducing method based on multiple DSPs |
Non-Patent Citations (1)
| Title |
|---|
| PUYANG WANG: "Generating High Quality Visible Images from SAR Images Using CNNs", 《ARXIV:1802.10036V1》 * |
Cited By (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109658344A (en) * | 2018-11-12 | 2019-04-19 | 哈尔滨工业大学（深圳） | Image de-noising method, device, equipment and storage medium based on deep learning |
| CN109658344B (en) * | 2018-11-12 | 2022-10-25 | 哈尔滨工业大学（深圳） | Image denoising method, device and equipment based on deep learning and storage medium |
| CN109636742B (en) * | 2018-11-23 | 2020-09-22 | 中国人民解放军空军研究院航空兵研究所 | Mode conversion method of SAR image and visible light image based on countermeasure generation network |
| CN109636742A (en) * | 2018-11-23 | 2019-04-16 | 中国人民解放军空军研究院航空兵研究所 | The SAR image of network and the mode conversion method of visible images are generated based on confrontation |
| CN109859147A (en) * | 2019-03-01 | 2019-06-07 | 武汉大学 | A kind of true picture denoising method based on generation confrontation network noise modeling |
| CN109859147B (en) * | 2019-03-01 | 2021-05-04 | 武汉大学 | Real image denoising method based on generation of antagonistic network noise modeling |
| CN110163801A (en) * | 2019-05-17 | 2019-08-23 | 深圳先进技术研究院 | A kind of Image Super-resolution and color method, system and electronic equipment |
| CN110163801B (en) * | 2019-05-17 | 2021-07-20 | 深圳先进技术研究院 | Image super-resolution and coloring method, system and electronic equipment |
| CN110751630A (en) * | 2019-09-30 | 2020-02-04 | 山东信通电子股份有限公司 | Power transmission line foreign matter detection method and device based on deep learning and medium |
| 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 |
| CN111862253A (en) * | 2020-07-14 | 2020-10-30 | 华中师范大学 | Sketch coloring method and system for generating confrontation network based on deep convolution |
| CN111862253B (en) * | 2020-07-14 | 2023-09-15 | 华中师范大学 | Sketch coloring method and system for generating countermeasure network based on deep convolution |
| CN115546351A (en) * | 2022-12-02 | 2022-12-30 | 耕宇牧星(北京)空间科技有限公司 | Convolution network-based synthetic aperture radar image coloring method |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108717698A (en) | A kind of high quality graphic generation method generating confrontation network based on depth convolution | |
| Lee et al. | Polarimetric SAR speckle filtering and the extended sigma filter | |
| Antropov et al. | Land cover and soil type mapping from spaceborne PolSAR data at L-band with probabilistic neural network | |
| CN111273238B (en) | SAR (synthetic aperture radar) wide-band and narrow-band interference simultaneous inhibition method based on low-rank recovery | |
| CN110516728B (en) | Polarized SAR terrain classification method based on denoising convolutional neural network | |
| Cetin et al. | SAR imaging from partial-aperture data with frequency-band omissions | |
| Choe et al. | Detection of oyster habitat in tidal flats using multi-frequency polarimetric SAR data | |
| CN109919870B (en) | SAR image speckle suppression method based on BM3D | |
| CN104076360B (en) | The sparse target imaging method of two-dimensional SAR based on compressed sensing | |
| CN110221256A (en) | SAR disturbance restraining method based on depth residual error network | |
| CN108133468B (en) | Constant false alarm rate ship detection method based on adaptive parameter enhancement and trail auxiliary detection | |
| CN103020922A (en) | PCA (principal component analysis) transformation based SAR (synthetic aperture radar) image speckle suppression method | |
| CN113655478B (en) | Imaging method and device | |
| Kabiri et al. | Detecting coral bleaching, using QuickBird multi-temporal data: a feasibility study at Kish Island, the Persian Gulf | |
| CN108196253A (en) | A kind of azimuth ambiguity suppression method for synthetic aperture radar | |
| CN111861905A (en) | SAR image speckle noise suppression method based on Gamma-Lee filtering | |
| Wang et al. | Application of mixture regression for improved polarimetric SAR speckle filtering | |
| Yang et al. | Adaptive coherency matrix estimation for polarimetric SAR imagery based on local heterogeneity coefficients | |
| Tao et al. | Radio frequency interference signature detection in radar remote sensing image using semantic cognition enhancement network | |
| CN110378896B (en) | TomosAR vegetation pest and disease damage monitoring method and device based on polarization coherence | |
| CN117115669B (en) | Object-level ground object sample self-adaptive generation method and system with double-condition quality constraint | |
| CN111398955B (en) | SAR image sidelobe removing method based on generation of antagonistic neural network | |
| Tuuk et al. | Compressed sensing radar amid noise and clutter using interference covariance information | |
| CN112989940A (en) | Raft culture area extraction method based on high-resolution three-satellite SAR image | |
| Luo et al. | Despeckling multitemporal polarimetric SAR data based on tensor decomposition |
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 | ||
| WW01 | Invention patent application withdrawn after publication |
Application publication date: 20181030 |
|
| WW01 | Invention patent application withdrawn after publication |