CN109410146A - A kind of image deblurring algorithm based on Bi-Skip-Net - Google Patents
A kind of image deblurring algorithm based on Bi-Skip-Net Download PDFInfo
- Publication number
- CN109410146A CN109410146A CN201811298475.8A CN201811298475A CN109410146A CN 109410146 A CN109410146 A CN 109410146A CN 201811298475 A CN201811298475 A CN 201811298475A CN 109410146 A CN109410146 A CN 109410146A
- Authority
- CN
- China
- Prior art keywords
- features
- skip
- feature
- image
- shallow
- 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 claims abstract description 23
- 238000005070 sampling Methods 0.000 claims description 12
- 230000009467 reduction Effects 0.000 claims description 8
- 208000037170 Delayed Emergence from Anesthesia Diseases 0.000 claims description 5
- 238000013461 design Methods 0.000 claims description 5
- 230000006835 compression Effects 0.000 claims description 4
- 238000007906 compression Methods 0.000 claims description 4
- 238000011176 pooling Methods 0.000 claims description 3
- 230000000694 effects Effects 0.000 abstract description 14
- 238000013135 deep learning Methods 0.000 abstract description 9
- 238000012545 processing Methods 0.000 abstract description 5
- 238000011084 recovery Methods 0.000 description 6
- 238000010586 diagram Methods 0.000 description 5
- 230000006870 function Effects 0.000 description 5
- 230000007246 mechanism Effects 0.000 description 5
- 230000003044 adaptive effect Effects 0.000 description 2
- 238000013528 artificial neural network Methods 0.000 description 2
- 239000000284 extract Substances 0.000 description 2
- 230000008569 process Effects 0.000 description 2
- 239000000654 additive Substances 0.000 description 1
- 230000000996 additive effect Effects 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 230000006399 behavior Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000002068 genetic effect Effects 0.000 description 1
- 238000003384 imaging method Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 238000003909 pattern recognition Methods 0.000 description 1
- 238000007781 pre-processing Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000011218 segmentation Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 238000012360 testing method Methods 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
- G06T5/00—Image enhancement or restoration
- G06T5/73—Deblurring; Sharpening
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Image Analysis (AREA)
- Image Processing (AREA)
Abstract
The present invention relates to digital image processing fields, especially a kind of image deblurring method based on Bi-Skip-Net, it is a kind of Bi-Skip-Net network to realize blur image restoration, it is intended to the problems such as solving high time complexity existing for existing deep learning deblurring algorithm, reconstruction inaccuracy, restored image there are grid effects.A kind of Bi-Skip-Net network disclosed by the invention as GAN (Generative Adversarial Network) generation network, aim to solve the problem that existing deep learning deblurring algorithm there are the shortcomings that, by comparing existing optimal algorithm, the present invention improves 0.1s on time complexity, can go up in image complex pattern originality and averagely improve 1dB.
Description
Technical Field
The invention relates to the field of digital image processing, in particular to a Bi-Skip-Net-based image deblurring method, which realizes blurred image restoration through a Bi-Skip-Net network.
Technical Field
Deblurring techniques are the subject of extensive research in the field of image and video processing. The blurring caused by camera shake seriously affects the imaging quality and visual impression of the image in a certain sense. As one of the important branches of the image preprocessing field, the improvement of the deblurring technology directly influences the performance of other computer vision algorithms, such as foreground segmentation, object detection, behavior analysis and the like; it also affects the coding performance of the picture. Therefore, it is imperative to develop a high performance deblurring algorithm.
Documents 1 to 3 describe a deblurring technique for image and video processing, a deep learning deblurring algorithm; document 1: kupyn O, Budzan V, Mykhailych M, et al, Deblurgan: Blind Motion Debluring Using conditional additive Networks [ J ]. arXiv preprint arXiv:1711.07064,2017. Document 2: NaH S, Kim T H, Lee K M. deep multi-scale volumetric neural network for dynamic scene deblocking [ C ]// CVPR.2017,1(2): 3. Document 3: sun J, Cao W, Xu Z, equivalent.left a relational neural network for non-uniform motion blur [ C ]// Proceedings of the IEEE Conference on Computer Vision and Pattern recognition.2015: 769-.
In general, image deblurring algorithms can be divided into conventional algorithms based on probabilistic models and deblurring algorithms based on deep learning. The traditional algorithm adopts a convolution model to explain the blur cause, and the process of camera shake can be mapped to a blur kernel trajectory PSF (Point Spread function). The method is characterized in that a clear image is restored under the condition that a fuzzy kernel is unknown, and the problem belongs to the ill-posed problem, so that the fuzzy kernel needs to be estimated firstly in the general sense, and then the restored image is obtained by carrying out deconvolution operation by utilizing the estimated fuzzy kernel. The deblurring algorithm based on deep learning utilizes a deep network structure to acquire potential information of an image, and further fuzzy image restoration is achieved. The deep learning deblurring algorithm can realize two operations of fuzzy kernel estimation and non-blind deconvolution to restore the image, and meanwhile, a countermeasure mechanism can be adopted to restore the image. This patent aims at solving the disadvantages of deblurring algorithms:
1) the time complexity is high, and the time complexity is high,
2) the recovery of the texture is not accurate and,
3) the restored image has a checkered effect.
Disclosure of Invention
The invention provides a Bi-Skip-Net network as a generating network of a GAN (generic adaptive network), aiming at solving the defects of the existing deep learning deblurring algorithm. Compared with the existing optimal algorithm, the time complexity of the method is improved by 0.1s, and the original performance of the image recovery image is improved by 1dB on average.
The technical scheme provided by the invention is as follows: (Note: the technical solution should be explained by natural language, but cannot be described by the way of description such as "figure", and the technical solution of the method is preferably written in the way of: step 1: step 2 …)
The invention adopts a generation countermeasure network mechanism to realize the restoration of the blurred image and designs a Bi-Skip-Net network as a generator in the network. The method comprises the following specific steps:
1): inputting a blurred image, and obtaining shallow layer characteristics through a convolution layer with convolution kernel size of 7x7 and step length of 1;
2): obtaining depth characteristics of the shallow layer characteristics under the current scale through 3 residual blocks;
3): carrying out a mode of downsampling and adding residual error on the depth features to obtain shallow features under the next scale;
4): repeating the steps 2 and 3 to obtain shallow layer features and depth features under different scales according to the specified down-sampling times n, and not obtaining the depth features under the minimum scale;
5): taking the shallow feature with the minimum scale as a basic feature;
6): the shallow feature of the previous scale is processed by a convolution layer with the convolution kernel size of 1x1 and the step length of 1 to obtain a shallow dimension reduction feature; the corresponding depth features are subjected to convolution layers with convolution kernel size of 3x3 and step length of 2 to obtain depth dimension reduction features, and the depth dimension reduction features and the basic features are connected in series to perform upsampling; connecting the up-sampled features and the shallow dimensionality reduction features in series to obtain basic features under the current scale;
7): repeating the step 6 until the sampling operation is cut off;
8): the obtained basic features are subjected to convolution layers with convolution kernel size of 7x7 and step length of 1 to obtain residual error features;
9): adding the residual error characteristics and the input image to obtain a restored image;
…
the Bi-Skip-Net plus residual mode is adopted as a generator.
In step 4), the predetermined number of downsampling times is 5.
The fuzzy image is used for obtaining a restored image through a generator, and the task of distinguishing the restored image from a clear image is as far as possible; and the task of the generator is to fool the discriminator as much as possible to reduce the ability to distinguish between the two images.
The Bi-Skip-Net network consists of three parts: a contact path (D), a Skip path (S), and an expand path (U). The Contract layer carries out downsampling to realize feature compression, the Skip layer is used for connecting deep features and shallow features, and the expand layer carries out upsampling. Wherein D, S, U are features corresponding to the down-sampling scale.
In the feature operation under the sampling scale, in a contact path, the current feature obtains a deep feature through 3 residual blocks (3xResBlock), and a residual mode of pooling (posing) and convolution addition is adopted to obtain the feature of the next scale; in the Skip path, the shallow features are compressed by convolution of 1x1, and the depth features are compressed by convolution of 3x 3; in the expand path, feature concatenation is achieved by concat, and feature upsampling is achieved by deconvolution of 3x 3.
The invention has the following technical effects: compared with the prior art, the invention adopts the Bi-Skip-Net network as the generating network of the GAN (genetic adaptive network), and has the following technical effects:
1. the time complexity is low; compared with the traditional method, the traditional motion blur removing method adopts two steps of fuzzy kernel estimation and non-blind deconvolution, and the two steps need to be iterated for multiple times to achieve a better recovery effect, so that the time for processing a single motion blur image is longer; the model designed by the invention can avoid time loss caused by multiple iterative optimization.
2. The texture is recovered accurately; compared with the traditional method, in the traditional method, inaccurate estimation of the fuzzy core can cause error recovery of image information in the recovery process, and non-blind deconvolution operation can often cause ringing effect on texture parts; the double-span connection network designed by the invention extracts the depth feature and the shallow feature on each layer of scale, and the network can recover more detailed information to a certain extent through feature connection.
3. Compared with the existing deep learning method, most of the existing deep learning methods are realized by adopting deconvolution layers in the up-sampling process, and each deconvolution has a certain sawtooth effect, so that the final restored image also has a plurality of sawteeth, namely the checkered effect provided by the invention.
For a better understanding of the principles and concepts of the invention, reference should be made to the following detailed description of the invention taken in conjunction with the accompanying drawings and examples. The description of specific embodiments does not limit the scope of the invention in any way.
Drawings
FIG. 1 is a diagram of the present invention of a generate countermeasure network mechanism;
FIG. 2 is a diagram of a Bi-Skip-Net network according to the present invention;
FIG. 3 illustrates the characteristic operation at one sampling scale of the present invention;
FIG. 4 Generator design: Bi-Skip-Net + residual error;
FIGS. 5a-d are subjective comparisons of the present invention with other algorithms; wherein,
FIG. 5 a: blurring the image;
FIG. 5 b: nah et al;
FIG. 5 c: the restorative effect of Kupyn et al;
FIG. 5 d: the restoration effect of Bi-Skip-Net.
Detailed Description
Fig. 1 is a diagram of a generation countermeasure network mechanism employed by the present invention. The fuzzy image is used for obtaining a restored image through a generator, and the task of distinguishing the restored image from a clear image is as far as possible; and the task of the generator is to fool the discriminator as much as possible to reduce the ability to distinguish between the two images.
The embodiment of the invention comprises the following specific steps:
(1) designing a generator and a discriminator, wherein the principle is as shown in FIG. 4, a fuzzy image of a building is generated by a Bi-Skip-Net generator, and a clear building picture is obtained; any other blurred image can be used to generate a sharp picture using this model.
(2) The network is trained using a loss function as follows,
whereinIn order to combat the loss-function,is the conditional loss function, and λ is the weight of the conditional loss function.
By maximisingTo optimize the discriminator D;
generator G is optimized by minimizing equation 3;
whereinThe design is as follows:
wherein, L and S respectively represent the output and truth value of the model at different levels, α takes the value of 1 or 2, and the whole condition loss function is normalized by the number c of channels, the width w and the height h.
(3) And taking the trained network as a final recovery model.
As shown in fig. 1, the method of the embodiment of the present invention employs a generation countermeasure network mechanism to achieve blurred image restoration. Fig. 2 is a structure diagram of a Bi-Skip-Net network, and a Bi-Skip-Net network is designed as a generator in the network structure shown in fig. 2.
The discriminator parameters in the structure diagram of the Bi-Skip-Net network are shown in a discriminator parameter table 1.
TABLE 1 discriminator parameter table
| # | Layer(s) | Dimension of parameter | Step size |
| 1 | conv | 32x3x5x5 | 2 |
| 2 | conv | 64x32x5x5 | 1 |
| 3 | conv | 64x64x5x5 | 2 |
| 4 | conv | 128x64x5x5 | 1 |
| 5 | conv | 128x128x5x5 | 4 |
| 6 | conv | 256x128x5x5 | 1 |
| 7 | conv | 256x256x5x5 | 4 |
| 8 | conv | 512x256x5x5 | 1 |
| 9 | conv | 512x512x4x4 | 4 |
| 10 | fc | 512x1x1x1 | - |
As shown in FIG. 2, the Bi-Skip-Net network designed by the embodiment of the present invention is composed of three parts: a contact path (D) including D0, D1, D2, and D3; a Skip path (S) including S0, S1, S2, and S3; and an expand path (U) including U0, U1, U2, and U3. The Contract layer carries out downsampling to realize feature compression, the Skip layer is used for connecting deep features and shallow features, and the expand layer carries out upsampling. Wherein D (D0, D1, D2 and D3), S (S0, S1, S2 and S3), and U (U0, U1, U2 and U3) are features at corresponding down-sampling scales.
Fig. 3 is a feature operation at one sampling scale, as shown in fig. 3, in a extract path, i.e., a compression path, a current feature obtains a deep feature through 3 residual blocks (3xResBlock), and a residual mode of pooling (stacking) and convolution addition is adopted to obtain a feature at a next scale; in the Skip path, namely the cross-connection path, the shallow features are compressed by convolution of 1x1, and the deep features are compressed by convolution of 3x 3; in the expand path, feature concatenation is achieved by concat, i.e., concatenate, and feature upsampling is achieved by deconvolution of 3x 3.
FIG. 4 Generator design: and the Bi-Skip-Net + residual takes a Bi-Skip-Net plus residual mode as a generator finally, as shown in FIG. 4.
The results of comparison of the implementation of the present invention with other algorithms are detailed in table 2. the present invention is compared with other algorithms for testing on GoPro data sets.
TABLE 2 comparison of the present invention with other algorithms tested on the GoPro dataset
Fig. 5a-d are subjective comparisons of the present invention with other algorithms. FIG. 5a is a blurred image, FIG. 5b is the restoration effect of Nah et al, FIG. 5c is the restoration effect of Kupyn et al, and FIG. 5d is the restoration effect of Bi-Skip-Net according to the present invention. The character "HARDWARE" at the lower left corner of the picture can not be identified or can not be identified in other three pictures, and the invention can clearly restore and identify. The subjective comparison of people shows that the method has obvious repairing effect on the blurred image.
It is noted that the disclosed embodiments are intended to aid in further understanding of the invention, but those skilled in the art will appreciate that: various substitutions and modifications are possible without departing from the spirit and scope of the invention and appended claims. Therefore, the invention should not be limited to the embodiments disclosed, but the scope of the invention is defined by the appended claims.
Claims (5)
1. An image deblurring method based on Bi-Skip-Net comprises the following steps:
1) inputting a blurred image, and obtaining shallow layer characteristics through a convolution layer with convolution kernel size of 7x7 and step length of 1;
2) obtaining depth characteristics of the shallow layer characteristics under the current scale through 3 residual blocks;
3) carrying out a mode of downsampling and adding residual error on the depth features to obtain shallow features under the next scale;
4) repeating the steps 2 and 3 to obtain shallow layer features and depth features under different scales according to the specified down-sampling times n, and not obtaining the depth features under the minimum scale;
5) taking the shallow feature with the minimum scale as a basic feature;
6) the shallow feature of the previous scale is processed by a convolution layer with the convolution kernel size of 1x1 and the step length of 1 to obtain a shallow dimension reduction feature; the corresponding depth features are subjected to convolution layers with convolution kernel size of 3x3 and step length of 2 to obtain depth dimension reduction features, and the depth dimension reduction features and the basic features are connected in series to perform upsampling; connecting the up-sampled features and the shallow dimensionality reduction features in series to obtain basic features under the current scale;
7) repeating the step 6 until the sampling operation is cut off;
8) the obtained basic features are subjected to convolution layers with convolution kernel size of 7x7 and step length of 1 to obtain residual error features;
9) adding the residual error characteristics and the input image to obtain a restored image;
10) and adopting a Bi-Skip-Net plus residual mode as a generator.
2. The image deblurring method of claim 1, wherein:
step 4) the number of downsampling times is 5.
3. The image deblurring method of claim 1, wherein:
the Bi-Skip-Net network consists of three parts: a contact path (D), a Skip path (S) and an expand path (U); the Contract layer carries out downsampling to realize feature compression, the Skip layer is used for connecting deep features and shallow features, and the expand layer carries out upsampling; wherein D, S, U are features corresponding to the down-sampling scale.
4. The image deblurring method of claim 3, wherein:
in the feature operation under the sampling scale, in a contact path, the current feature obtains a deep feature through 3 residual blocks (3xResBlock), and a residual mode of pooling (posing) and convolution addition is adopted to obtain the feature of the next scale; in the Skip path, the shallow features are compressed by convolution of 1x1, and the depth features are compressed by convolution of 3x 3; in the expand path, feature concatenation is achieved by concat, and feature upsampling is achieved by deconvolution of 3x 3.
5. The image deblurring method of claim 1, wherein:
the generator of step 10) is designed in such a way that,
① the network is trained using the following loss function,
whereinIn order to combat the loss-function,is a conditional loss function, and λ is the weight of the conditional loss function;
by maximisingTo optimize the discriminator D;
generator G is optimized by minimizing equation 3;
whereinThe design is as follows:
wherein, L and S respectively represent the output and truth value of the model at different levels, α takes the value as 1 or 2, and the whole condition loss function is normalized by the number c of channels, the width w and the height h;
②, the trained network is used as the final restoration model.
Priority Applications (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811298475.8A CN109410146A (en) | 2018-11-02 | 2018-11-02 | A kind of image deblurring algorithm based on Bi-Skip-Net |
| PCT/CN2018/117634 WO2020087607A1 (en) | 2018-11-02 | 2018-11-27 | Bi-skip-net-based image deblurring method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811298475.8A CN109410146A (en) | 2018-11-02 | 2018-11-02 | A kind of image deblurring algorithm based on Bi-Skip-Net |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109410146A true CN109410146A (en) | 2019-03-01 |
Family
ID=65471437
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811298475.8A Pending CN109410146A (en) | 2018-11-02 | 2018-11-02 | A kind of image deblurring algorithm based on Bi-Skip-Net |
Country Status (2)
| Country | Link |
|---|---|
| CN (1) | CN109410146A (en) |
| WO (1) | WO2020087607A1 (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110570375A (en) * | 2019-09-06 | 2019-12-13 | 腾讯科技（深圳）有限公司 | image processing method, image processing device, electronic device and storage medium |
| CN111612711A (en) * | 2019-05-31 | 2020-09-01 | 北京理工大学 | Improved picture deblurring method based on generation countermeasure network |
| CN112102184A (en) * | 2020-09-04 | 2020-12-18 | 西北工业大学 | Image deblurring method based on Scale-Encoder-Decoder-Net network |
| CN113570516A (en) * | 2021-07-09 | 2021-10-29 | 湖南大学 | Image blind motion deblurring method based on CNN-Transformer hybrid self-encoder |
Families Citing this family (20)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111986102B (en) * | 2020-07-15 | 2024-02-27 | 万达信息股份有限公司 | Digital pathological image deblurring method |
| CN112070658B (en) * | 2020-08-25 | 2024-04-16 | 西安理工大学 | Deep learning-based Chinese character font style migration method |
| CN112070693B (en) * | 2020-08-27 | 2024-03-26 | 西安理工大学 | Single dust image recovery method based on gray world adaptive network |
| CN112184590B (en) * | 2020-09-30 | 2024-03-26 | 西安理工大学 | Single dust image recovery method based on gray world self-guiding network |
| CN112330554B (en) * | 2020-10-30 | 2024-01-19 | 西安工业大学 | Structure learning method for deconvolution of astronomical image |
| CN112561819A (en) * | 2020-12-17 | 2021-03-26 | 温州大学 | Self-filtering image defogging algorithm based on self-supporting model |
| CN112634163B (en) * | 2020-12-29 | 2024-10-15 | 南京大学 | Method for removing image motion blur based on improved cyclic generation countermeasure network |
| CN113538263A (en) * | 2021-06-28 | 2021-10-22 | 江苏威尔曼科技有限公司 | Motion blur removing method, medium, and device based on improved DeblurgAN model |
| CN113592736B (en) * | 2021-07-27 | 2024-01-12 | 温州大学 | Semi-supervised image deblurring method based on fused attention mechanism |
| CN113610721B (en) * | 2021-07-27 | 2024-09-24 | 河南大学 | Image restoration method for generating countermeasure network based on partial convolution |
| CN113947589B (en) * | 2021-10-26 | 2024-08-02 | 北京理工大学 | Missile-borne image deblurring method based on countermeasure generation network |
| CN114119395B (en) * | 2021-11-15 | 2024-06-11 | 北京理工大学 | Image processing system and method integrating distortion detection and restoration |
| CN114240771A (en) * | 2021-11-23 | 2022-03-25 | 无锡学院 | Image deblurring system and method based on dual control network |
| CN114511465B (en) * | 2022-02-21 | 2024-08-20 | 华东交通大学 | Image restoration method and system based on improvement DCGAN |
| CN114723630B (en) * | 2022-03-31 | 2024-09-06 | 福州大学 | Image deblurring method and system based on cavity double-residual multi-scale depth network |
| CN114936977A (en) * | 2022-05-31 | 2022-08-23 | 东南大学 | Image deblurring method based on channel attention and cross-scale feature fusion |
| CN114913095B (en) * | 2022-06-08 | 2024-03-12 | 西北工业大学 | Depth deblurring method based on domain adaptation |
| CN114841897B (en) * | 2022-06-08 | 2024-03-15 | 西北工业大学 | Depth deblurring method based on self-adaptive fuzzy kernel estimation |
| CN115760589A (en) * | 2022-09-30 | 2023-03-07 | 浙江大学 | Image optimization method and device for motion blurred image |
| CN117058038B (en) * | 2023-08-28 | 2024-04-30 | 北京航空航天大学 | Diffraction blurred image restoration method based on even convolution deep learning |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106251303A (en) * | 2016-07-28 | 2016-12-21 | 同济大学 | A kind of image denoising method using the degree of depth full convolutional encoding decoding network |
| US20170365046A1 (en) * | 2014-08-15 | 2017-12-21 | Nikon Corporation | Algorithm and device for image processing |
| CN107689034A (en) * | 2017-08-16 | 2018-02-13 | 清华-伯克利深圳学院筹备办公室 | A kind of training method of neutral net, denoising method and device |
| CN108629743A (en) * | 2018-04-04 | 2018-10-09 | 腾讯科技（深圳）有限公司 | Processing method, device, storage medium and the electronic device of image |
Family Cites Families (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8737687B2 (en) * | 2011-12-30 | 2014-05-27 | Honeywell International Inc. | System and method for tracking a subject using raw images and tracking errors |
| CN108460742A (en) * | 2018-03-14 | 2018-08-28 | 日照职业技术学院 | A kind of image recovery method based on BP neural network |
| CN108711141B (en) * | 2018-05-17 | 2022-02-15 | 重庆大学 | Motion blurred image blind restoration method using improved generation type countermeasure network |
-
2018
- 2018-11-02 CN CN201811298475.8A patent/CN109410146A/en active Pending
- 2018-11-27 WO PCT/CN2018/117634 patent/WO2020087607A1/en active Application Filing
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20170365046A1 (en) * | 2014-08-15 | 2017-12-21 | Nikon Corporation | Algorithm and device for image processing |
| CN106251303A (en) * | 2016-07-28 | 2016-12-21 | 同济大学 | A kind of image denoising method using the degree of depth full convolutional encoding decoding network |
| CN107689034A (en) * | 2017-08-16 | 2018-02-13 | 清华-伯克利深圳学院筹备办公室 | A kind of training method of neutral net, denoising method and device |
| CN108629743A (en) * | 2018-04-04 | 2018-10-09 | 腾讯科技（深圳）有限公司 | Processing method, device, storage medium and the electronic device of image |
Non-Patent Citations (3)
| Title |
|---|
| KAI LIN 等: "Adaptive Integration Skip Compensation Neural Networks for Removing Mixed Noise in Image", 《PCM2018:ADVANCES IN MULTIMEDIA INFORMATION PROCESSING》 * |
| OREST KUPYN 等: "DeblurGAN: Blind Motion Deblurring Using Conditional Adversarial Networks", 《ARXIV》 * |
| WEIWEI FU 等: "Arrears Prediction For Electricity Customer Through Wgan-Gp", 《2017 IEEE 2ND INFORMATION TECHNOLOGY, NETWORKING, ELECTRONIC AND AUTOMATION CONTROL CONFERENCE（ITNEC）》 * |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111612711A (en) * | 2019-05-31 | 2020-09-01 | 北京理工大学 | Improved picture deblurring method based on generation countermeasure network |
| CN111612711B (en) * | 2019-05-31 | 2023-06-09 | 北京理工大学 | Picture deblurring method based on generation of countermeasure network improvement |
| CN110570375A (en) * | 2019-09-06 | 2019-12-13 | 腾讯科技（深圳）有限公司 | image processing method, image processing device, electronic device and storage medium |
| CN110570375B (en) * | 2019-09-06 | 2022-12-09 | 腾讯科技（深圳）有限公司 | Image processing method, device, electronic device and storage medium |
| CN112102184A (en) * | 2020-09-04 | 2020-12-18 | 西北工业大学 | Image deblurring method based on Scale-Encoder-Decoder-Net network |
| CN113570516A (en) * | 2021-07-09 | 2021-10-29 | 湖南大学 | Image blind motion deblurring method based on CNN-Transformer hybrid self-encoder |
| CN113570516B (en) * | 2021-07-09 | 2022-07-22 | 湖南大学 | Image blind motion deblurring method based on CNN-Transformer hybrid self-encoder |
Also Published As
| Publication number | Publication date |
|---|---|
| WO2020087607A1 (en) | 2020-05-07 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109410146A (en) | A kind of image deblurring algorithm based on Bi-Skip-Net | |
| Anwar et al. | Densely residual laplacian super-resolution | |
| Zhang et al. | Image restoration: From sparse and low-rank priors to deep priors [lecture notes] | |
| Zhang et al. | One-two-one networks for compression artifacts reduction in remote sensing | |
| Yu et al. | A unified learning framework for single image super-resolution | |
| Sun et al. | Lightweight image super-resolution via weighted multi-scale residual network | |
| CN112435191B (en) | Low-illumination image enhancement method based on fusion of multiple neural network structures | |
| CN109544475A (en) | Bi-Level optimization method for image deblurring | |
| CN111553867B (en) | Image deblurring method and device, computer equipment and storage medium | |
| CN105590296B (en) | A kind of single-frame images Super-Resolution method based on doubledictionary study | |
| CN112036454A (en) | Image classification method based on multi-core dense connection network | |
| Anwar et al. | Attention-based real image restoration | |
| CN111489305B (en) | Image enhancement method based on reinforcement learning | |
| An et al. | RBDN: Residual bottleneck dense network for image super-resolution | |
| Zhou et al. | Sparse representation with enhanced nonlocal self-similarity for image denoising | |
| Hua et al. | Dynamic scene deblurring with continuous cross-layer attention transmission | |
| Ren et al. | A lightweight object detection network in low-light conditions based on depthwise separable pyramid network and attention mechanism on embedded platforms | |
| Xiao et al. | Frequency-Assisted Mamba for Remote Sensing Image Super-Resolution | |
| ZhiPing et al. | A new generative adversarial network for texture preserving image denoising | |
| Tiwari | A pattern classification based approach for blur classification | |
| CN104123707A (en) | Local rank priori based single-image super-resolution reconstruction method | |
| Wei et al. | Image denoising with deep unfolding and normalizing flows | |
| Kasem et al. | DRCS-SR: Deep robust compressed sensing for single image super-resolution | |
| Huang et al. | A non-local enhanced network for image restoration | |
| Nie et al. | Image restoration from patch-based compressed sensing measurement |
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: 20190301 |