CN113012069B - Optical remote sensing image quality improving method combining deep learning under wavelet transform domain - Google Patents
Optical remote sensing image quality improving method combining deep learning under wavelet transform domain Download PDFInfo
- Publication number
- CN113012069B CN113012069B CN202110285261.2A CN202110285261A CN113012069B CN 113012069 B CN113012069 B CN 113012069B CN 202110285261 A CN202110285261 A CN 202110285261A CN 113012069 B CN113012069 B CN 113012069B
- Authority
- CN
- China
- Prior art keywords
- remote sensing
- optical remote
- image
- wavelet transform
- sensing image
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 230000003287 optical effect Effects 0.000 title claims abstract description 61
- 238000000034 method Methods 0.000 title claims abstract description 35
- 238000013135 deep learning Methods 0.000 title claims abstract description 21
- 238000012545 processing Methods 0.000 claims abstract description 23
- 230000009466 transformation Effects 0.000 claims abstract description 21
- 238000012549 training Methods 0.000 claims abstract description 19
- 230000006870 function Effects 0.000 claims abstract description 11
- 238000013527 convolutional neural network Methods 0.000 claims description 3
- 238000010586 diagram Methods 0.000 description 6
- 238000013507 mapping Methods 0.000 description 5
- 238000003672 processing method Methods 0.000 description 4
- 230000006872 improvement Effects 0.000 description 3
- 238000001514 detection method Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 230000015556 catabolic process Effects 0.000 description 1
- 238000006731 degradation reaction Methods 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 238000003384 imaging method Methods 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 230000008569 process Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/70—Denoising; Smoothing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T3/00—Geometric image transformations in the plane of the image
- G06T3/40—Scaling of whole images or parts thereof, e.g. expanding or contracting
- G06T3/4053—Scaling of whole images or parts thereof, e.g. expanding or contracting based on super-resolution, i.e. the output image resolution being higher than the sensor resolution
-
- 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
-
- 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]
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D30/00—Reducing energy consumption in communication networks
- Y02D30/70—Reducing energy consumption in communication networks in wireless communication networks
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Biophysics (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Health & Medical Sciences (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses an optical remote sensing image quality improving method combining deep learning under a wavelet transform domain, which can be used for acquiring a high-quality optical remote sensing image and solves the problem that the high-quality optical remote sensing image cannot be acquired by adopting a denoising or super-resolution reconstruction technology at the same time when the remote sensing image is processed in the prior art. The method comprises the following steps: 1. in the training stage, the optical remote sensing image is subjected to wavelet transformation, denoising and super-resolution reconstruction at the same time, and then subjected to wavelet inverse transformation to obtain an optical remote sensing reconstruction image, and the optical remote sensing reconstruction image is subjected to discrimination processing by adopting a relative loss function to obtain ideal network parameters; 2. in the using stage, an ideal network parameter is used for initializing the system, the optical remote sensing image is subjected to wavelet transformation, denoising and super-resolution reconstruction, and then wavelet inverse transformation is carried out, so that an optical remote sensing reconstructed image is finally obtained, and the quality of the optical remote sensing image is further improved.
Description
Technical Field
The invention belongs to the field of optical remote sensing image processing, and particularly relates to an optical remote sensing image quality improving method combining deep learning under a wavelet transform domain, which can be used for high-quality acquisition of an optical remote sensing image.
Background
High quality optical remote sensing images are very useful in object detection, object recognition, and image classification. High quality optical remote sensing images are difficult to obtain due to the accuracy of the imaging equipment and the atmospheric environment. In the prior art, most of the techniques of denoising or Super-resolution reconstruction are adopted to acquire high-quality optical remote sensing images, such as ESRGAN image processing methods shown in FIG. 1 (see Wang X, yu K, wu S, et al ESRGAN: enhanced Super-Resolution Generative Adversarial Networks [ C ] European Conference on Computer vision Springer, cham, 2018.), and Super-resolution reconstruction of images can be realized based on deep learning, but denoising processing of images cannot be realized; the DnCNN image processing method shown in FIG. 2 (see Zhang K, zuo W, chen Y, et al Beyond a Gaussian Denoiser: residual Learning of Deep CNN for Image Denoising [ J ]. IEEE Transactions on Image Processing,2016, 26 (7): 3142-3155.) is a deep-layer image denoising method based on deep learning, and based on a residual network, batch normalization is combined to realize denoising processing on an image, but super-resolution reconstruction on the image cannot be realized; because the remote sensing image has a complex structure and large noise, the quality of the optical remote sensing image obtained by only adopting a denoising or super-resolution reconstruction method cannot meet the actual requirements, and further, the optical remote sensing image with higher quality cannot be obtained.
Disclosure of Invention
The invention provides an optical remote sensing image quality improving method combining deep learning under a wavelet transform domain, which can be used for any image type needing quality improvement, can simultaneously perform denoising and super-resolution reconstruction on an optical remote sensing image, and realizes acquisition of a high-quality optical remote sensing image.
The technical scheme of the invention is as follows: the optical remote sensing image quality improving method combining with deep learning under wavelet transform domain comprises the following steps:
step 1, training phase:
1.1 Performing wavelet transformation processing on the optical remote sensing images in the training set to obtain subband signals with different frequencies;
1.2 Carrying out denoising and super-resolution reconstruction processing on the frequency subband signals through a dense block network at the same time to obtain frequency subband reconstruction signals;
1.3 Performing wavelet inverse transformation on the frequency sub-band reconstruction signals to obtain an optical remote sensing reconstruction image;
1.4 The optical remote sensing reconstruction image and the corresponding high-quality truth image are subjected to discrimination processing, and the loss condition of the current cycle is obtained through the relative loss and total variation loss function operation of the generator module and the relative loss function operation of the discriminator module, and the cycle is ended;
1.5 Repeating the steps 1.1) -1.4) until the iteration is finished after all the optical remote sensing images in the training set are circulated once, and updating network parameters according to the comprehensive loss condition of the iteration;
1.6 Repeating the steps 1.1) -1.5) until the iteration times reach a set value, and obtaining ideal network parameters after training is finished;
step 2, use stage:
2.1 Initializing the system by using the ideal network parameters obtained in the training stage;
2.2 Performing wavelet transformation processing on the optical remote sensing image to be upgraded to obtain subband signals with different frequencies;
2.3 Carrying out denoising and super-resolution reconstruction processing on the frequency subband signals through a dense block network at the same time to obtain frequency subband reconstruction signals;
2.4 Performing wavelet inverse transformation on the frequency sub-band reconstruction signals to obtain an optical remote sensing reconstruction image.
Further, in step 1.2 and step 2.3, the dense block network adopts a global residual algorithm, a local residual algorithm and a dense connection algorithm.
Further, dense block networks use a bottleneck structure, comprising a 1 x 1 convolution kernel, a 3 x 3 convolution kernel, and a 1 x 1 convolution kernel.
Further, in step 1.1 and step 2.2, the wavelet transform is a haar wavelet transform; in step 1.3 and step 2.4, the inverse wavelet transform is a haar inverse wavelet transform.
Further, in step 1.1 and step 2.2, the frequency subband signals are the low frequency subband of the original image, the high frequency subband in the vertical direction, the high frequency subband in the horizontal direction and the high frequency subband signal in the diagonal direction.
Further, steps 1.1-1.4 are implemented using a generation countermeasure network or a conventional convolutional neural network.
Compared with the prior art, the invention has the following beneficial effects:
1) The optical remote sensing image quality improving method combining with the deep learning under the wavelet transform domain can simultaneously denoising and reconstructing the optical remote sensing image with super resolution, and realizes the acquisition of the high-quality optical remote sensing image.
2) The optical remote sensing image quality improving method combining deep learning under the wavelet transform domain expands the application range of the remote sensing image in target detection, target identification and image classification and provides an effective way for obtaining high-quality remote sensing images.
3) The optical remote sensing image quality improving method combining deep learning in the wavelet transform domain effectively reduces the acquisition cost of high-quality remote sensing images, and efficiently and rapidly realizes iteration and realization of the high-quality remote sensing images.
4) According to the optical remote sensing image quality improvement method combining deep learning under the wavelet transform domain, a generator model operates under the wavelet transform domain, and total variation loss is increased in a loss function so as to further enhance the reconstruction effect; meanwhile, the relative loss is used for replacing the traditional pixel-level loss by the discriminator model, so that the whole network is better converged, and finally, the optical remote sensing image reconstruction effect is improved.
Drawings
FIG. 1 is a schematic diagram of a conventional ESRGAN image processing method;
FIG. 2 is a schematic diagram of a conventional DnCNN image processing method;
FIG. 3 is a schematic diagram of a training stage architecture in the optical remote sensing image quality improvement method combining deep learning in the wavelet transform domain of the present invention;
FIG. 4 is a flow chart of a haar wavelet transform in the method of the present invention;
FIG. 5 is a schematic diagram of the principle of generating an countermeasure network in the method of the present invention;
FIG. 6 is a schematic diagram of the principle of the residual algorithm in the method of the present invention;
FIG. 7 is a schematic diagram of a dense block network in the method of the present invention.
Reference numerals illustrate:
the system comprises a 1-generator model, a 2-discriminator model, a 101-wavelet transform module, a 102-dense residual learning module and a 103-wavelet inverse transform module.
Detailed Description
The following describes specific embodiments of the present invention in detail with reference to the drawings.
As shown in fig. 3, the optical remote sensing image quality improving method combined with deep learning under the wavelet transform domain adopts a framework for generating an countermeasure network, and comprises a generator model 1 and a discriminator model 2. The generator model 1 comprises a wavelet transformation module 101, a dense residual learning module 102 and a wavelet inverse transformation module 103; the arbiter model 2 comprises a VGG-54 network.
The invention provides an optical remote sensing image quality improving method combining deep learning under a wavelet transform domain. The denoising and super-resolution reconstruction of the optical remote sensing image are fused into a unified frame, so that better image quality can be obtained. The specific process is as follows: in order to obtain more reconstruction detail information, firstly, a framework for generating an countermeasure network is adopted, and then, in order to enable a finally reconstructed image to have good performance on objective evaluation indexes such as peak signal-to-noise ratio, structural similarity and the like, a residual network (local residual network and global residual network) is combined in a generation part of the final reconstructed image. In the aspect of the loss function, the loss function of the generated part is combined with total variation loss, so that the reconstructed detail information is further enhanced; the discrimination section uses relative loss instead of the conventional discrimination loss calculation method, in order to make the whole network better converged. Finally, the different frequency components of the original image are performed separately under the wavelet transform domain.
Wavelet transformation has proven to be a very efficient image restoration method. The image can be transformed into a series of coefficients of the same size by wavelet transformation. And predicting the wavelet coefficient by using a sparse coding algorithm, and finally reconstructing the sub-band after the prediction detail to obtain a final high-quality optical image. The haar wavelet transform operation is shown in fig. 4, where the low frequency subbands of the original image represent a global topology, and the other three subbands represent high frequency portions in the vertical, horizontal and diagonal directions, respectively. In addition, with the inverse implementation of the wavelet transform, a final image of these four subband coefficients can be obtained. The four sub-band coefficients can be processed by a denoising and super-resolution reconstruction algorithm respectively, and finally the processed four sub-bands are subjected to Harr wavelet inverse transformation to obtain a final image.
As shown in fig. 5, the main structure of the present invention for generating an countermeasure generation network includes a generator model and a discriminator model. Wherein both the generator model 1 and the discriminant model 2 are convolutional neural networks.
The residual network solves the degradation problem according to the structure of fig. 6: the residual network contains an identity map and a residual map. Identity mapping refers to the "curved" portion of the graph, residual mapping refers to the remaining portion that is not "curved", and F (x) is the pre-summation network mapping.
The mechanism of the dense block network is more aggressive than the residual network: i.e. all layers are interconnected. That is, each layer will accept all its previous layers as its additional inputs, as shown in fig. 7, H (x) is the network map input to the summation. Conventional networks map x at the identity of layer i if formulated l The method comprises the following steps:
x l =H l (x l-1 ) …(2)
whereas for the residual network the identity mapping x from the upper layer input is increased l-1 :
x l =H l (x l-1 )+x l-1 …(3)
In a dense block network, all the previous layers are connected as inputs:
x l =H l ([x 0 ,x 1 ,...,x l-1 ]) …(4)
h in l (x l-1 ) Representing the identity mapping to layer I-1, H is performed for layer I l Performing function operation; h l ([x 0 ,x 1 ,。。。。。,x l-1 ]) Represents from layer 0 to layer l-Layer 1 output.
In the training stage of the invention, the generator model 1 and the discriminator model 2 both participate in the processing of the optical remote sensing image; in the use phase of the invention, only the generator model 1 participates in the processing of the optical remote sensing image. In the invention, two parallel channels of the residual network and the dense network finish denoising and super-resolution reconstruction processing on the image.
The training phase comprises the following steps:
1.1 Performing wavelet transformation processing on the optical remote sensing images in the training set to obtain subband signals with different frequencies; downsampling a high-quality truth image, and adding noise to obtain a training set;
1.2 Carrying out denoising and super-resolution reconstruction processing on the frequency subband signals through a dense block network at the same time to obtain frequency subband reconstruction signals;
1.3 Performing wavelet inverse transformation on the frequency sub-band reconstruction signals to obtain an optical remote sensing reconstruction image;
1.4 The optical remote sensing reconstruction image and the corresponding high-quality truth image are subjected to discrimination processing, and the loss condition of the current cycle is obtained through the relative loss and total variation loss function operation of the generator module and the relative loss function operation of the discriminator module, and the cycle is ended;
1.5 Repeating the steps 1.1) -1.4) until the iteration is finished after all the optical remote sensing images in the training set are circulated once, and updating network parameters according to the comprehensive loss condition of the iteration;
1.6 Repeating the steps 1.1) -1.5) until the iteration times reach a set value, and obtaining ideal network parameters after training is finished;
the use stage comprises the following steps:
step 1, initializing a system by using ideal network parameters obtained in a training stage;
step 2, performing wavelet transformation processing on the optical remote sensing image with the quality to be improved to obtain subband signals with different frequencies;
step 3, carrying out denoising and super-resolution reconstruction processing on the frequency subband signals through a dense block network at the same time to obtain frequency subband reconstruction signals;
and 4, carrying out wavelet inverse transformation on the frequency sub-band reconstruction signals to obtain an optical remote sensing reconstruction image.
The foregoing disclosure is merely illustrative of specific embodiments of the invention, but the embodiments are not limited thereto and variations within the scope of the invention will be apparent to those skilled in the art.
Claims (6)
1. The optical remote sensing image quality improving method combining deep learning under the wavelet transform domain is characterized by comprising the following steps of:
step 1, training phase:
1.1 Performing wavelet transformation processing on the optical remote sensing images in the training set to obtain subband signals with different frequencies;
1.2 Carrying out denoising and super-resolution reconstruction processing on the frequency subband signals through a dense block network at the same time to obtain frequency subband reconstruction signals;
1.3 Performing wavelet inverse transformation on the frequency sub-band reconstruction signals to obtain an optical remote sensing reconstruction image;
1.4 The optical remote sensing reconstruction image and the corresponding high-quality truth image are subjected to discrimination processing, and the loss condition of the current cycle is obtained through the relative loss and total variation loss function operation of the generator module and the relative loss function operation of the discriminator module, and the cycle is ended;
1.5 Repeating the steps 1.1) -1.4) until the iteration is finished after all the optical remote sensing images in the training set are circulated once, and updating network parameters according to the comprehensive loss condition of the iteration;
1.6 Repeating the steps 1.1) -1.5) until the iteration times reach a set value, and obtaining ideal network parameters after training is finished;
step 2, use stage:
2.1 Initializing the system by using the ideal network parameters obtained in the training stage;
2.2 Performing wavelet transformation processing on the optical remote sensing image to be upgraded to obtain subband signals with different frequencies;
2.3 Carrying out denoising and super-resolution reconstruction processing on the frequency subband signals through a dense block network at the same time to obtain frequency subband reconstruction signals;
2.4 Performing wavelet inverse transformation on the frequency sub-band reconstruction signals to obtain an optical remote sensing reconstruction image.
2. The method for improving the quality of an optical remote sensing image by combining deep learning under a wavelet transform domain according to claim 1, wherein the method comprises the following steps of:
in the step 1.2 and the step 2.3, the dense block network adopts a global residual algorithm, a local residual algorithm and a dense connection algorithm.
3. The method for improving the quality of an optical remote sensing image by combining deep learning under a wavelet transform domain according to claim 2, wherein the method comprises the following steps of: the dense block network uses a bottleneck structure comprising a 1 x 1 convolution kernel, a 3 x 3 convolution kernel, and a 1 x 1 convolution kernel.
4. The method for improving the quality of an optical remote sensing image combined with deep learning under a wavelet transform domain according to claim 1, 2 or 3, wherein the method comprises the following steps of:
in step 1.1 and step 2.2, the wavelet transform is a haar wavelet transform;
in step 1.3 and step 2.4, the inverse wavelet transform is a haar inverse wavelet transform.
5. The method for improving the quality of an optical remote sensing image by combining deep learning under a wavelet transform domain according to claim 4, wherein the method comprises the following steps:
in step 1.1 and step 2.2, the frequency subband signals are the low frequency subband of the original image, the high frequency subband in the vertical direction, the high frequency subband in the horizontal direction and the high frequency subband signal in the diagonal direction.
6. The method for improving the quality of an optical remote sensing image by combining deep learning under a wavelet transform domain according to claim 5, wherein the method comprises the following steps:
steps 1.1-1.4 are implemented using a generation countermeasure network or a conventional convolutional neural network.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110285261.2A CN113012069B (en) | 2021-03-17 | 2021-03-17 | Optical remote sensing image quality improving method combining deep learning under wavelet transform domain |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110285261.2A CN113012069B (en) | 2021-03-17 | 2021-03-17 | Optical remote sensing image quality improving method combining deep learning under wavelet transform domain |
Publications (2)
Publication Number | Publication Date |
---|---|
CN113012069A CN113012069A (en) | 2021-06-22 |
CN113012069B true CN113012069B (en) | 2023-09-05 |
Family
ID=76409100
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202110285261.2A Active CN113012069B (en) | 2021-03-17 | 2021-03-17 | Optical remote sensing image quality improving method combining deep learning under wavelet transform domain |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN113012069B (en) |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108537742A (en) * | 2018-03-09 | 2018-09-14 | 天津大学 | A kind of panchromatic sharpening method of remote sensing images based on generation confrontation network |
CN108564109A (en) * | 2018-03-21 | 2018-09-21 | 天津大学 | A kind of Remote Sensing Target detection method based on deep learning |
Family Cites Families (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US11422217B2 (en) * | 2019-06-26 | 2022-08-23 | Siemens Healthcare Gmbh | Progressive generative adversarial network in medical image reconstruction |
-
2021
- 2021-03-17 CN CN202110285261.2A patent/CN113012069B/en active Active
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108537742A (en) * | 2018-03-09 | 2018-09-14 | 天津大学 | A kind of panchromatic sharpening method of remote sensing images based on generation confrontation network |
CN108564109A (en) * | 2018-03-21 | 2018-09-21 | 天津大学 | A kind of Remote Sensing Target detection method based on deep learning |
Non-Patent Citations (1)
Title |
---|
毕晓君 ; 潘梦迪 ; .基于生成对抗网络的机载遥感图像超分辨率重建.智能系统学报.2020,(第01期),全文. * |
Also Published As
Publication number | Publication date |
---|---|
CN113012069A (en) | 2021-06-22 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Phung et al. | Wavelet diffusion models are fast and scalable image generators | |
CN109035142B (en) | Satellite image super-resolution method combining countermeasure network with aerial image prior | |
CN106780338B (en) | Rapid super-resolution reconstruction method based on anisotropy | |
Min et al. | Blind deblurring via a novel recursive deep CNN improved by wavelet transform | |
Xiu et al. | Composite convolutional neural network for noise deduction | |
Manimala et al. | Convolutional neural network for sparse reconstruction of MR images interposed with gaussian noise | |
Cherian et al. | A Novel AlphaSRGAN for Underwater Image Super Resolution. | |
CN114998107A (en) | Image blind super-resolution network model, method, equipment and storage medium | |
Thakur et al. | Gradient and multi scale feature inspired deep blind gaussian denoiser | |
Wu et al. | Dcanet: Dual convolutional neural network with attention for image blind denoising | |
CN113509165B (en) | Complex rapid magnetic resonance imaging method based on CAR2UNet network | |
CN117726540A (en) | Image denoising method for enhanced gate control converter | |
CN113012069B (en) | Optical remote sensing image quality improving method combining deep learning under wavelet transform domain | |
Song et al. | Unsupervised denoising for satellite imagery using wavelet subband cyclegan | |
Balasubramanian et al. | MRI super-resolution using generative adversarial network and discrete wavelet transform | |
Kuzhali et al. | Automated image denoising model: Contribution towards optimized internal and external basis | |
CN112927169A (en) | Remote sensing image denoising method based on wavelet transformation and improved weighted nuclear norm minimization | |
Yu et al. | SAR image despeckling based on U-shaped transformer from a single noisy image | |
Li et al. | Visual encoding and decoding of the human brain based on shared features | |
Liu et al. | Synthetic aperture radar image despeckling using convolutional neural networks in wavelet domain | |
Gao et al. | High-quality image interpolation via local autoregressive and nonlocal 3-D sparse regularization | |
Tsai et al. | Unsupervised cross domain learning for noise removal from a single image | |
CN115797183B (en) | Image super-resolution reconstruction method | |
Alimagadov et al. | Wavelet-Based Loss Function for Image Denoising by Neural Networks | |
Christopher et al. | ADMM based Deep Denoiser Prior for Enhancing Single Coil Magnitude MR images |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |