CN109410149A - A kind of CNN denoising method extracted based on Concurrent Feature - Google Patents
A kind of CNN denoising method extracted based on Concurrent Feature Download PDFInfo
- Publication number
- CN109410149A CN109410149A CN201811323811.XA CN201811323811A CN109410149A CN 109410149 A CN109410149 A CN 109410149A CN 201811323811 A CN201811323811 A CN 201811323811A CN 109410149 A CN109410149 A CN 109410149A
- Authority
- CN
- China
- Prior art keywords
- cnn
- denoising
- image
- network model
- concurrent feature
- 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.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 50
- 238000012549 training Methods 0.000 claims abstract description 20
- 238000000605 extraction Methods 0.000 claims abstract description 4
- 238000013178 mathematical model Methods 0.000 claims description 7
- 230000004927 fusion Effects 0.000 claims description 6
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 claims description 4
- 230000004913 activation Effects 0.000 claims description 3
- 239000011248 coating agent Substances 0.000 claims description 3
- 238000000576 coating method Methods 0.000 claims description 3
- 238000013527 convolutional neural network Methods 0.000 description 45
- 230000000694 effects Effects 0.000 description 13
- 238000012360 testing method Methods 0.000 description 10
- 230000006870 function Effects 0.000 description 8
- 239000000284 extract Substances 0.000 description 6
- 230000013016 learning Effects 0.000 description 6
- 238000002474 experimental method Methods 0.000 description 5
- 238000001914 filtration Methods 0.000 description 5
- 238000013507 mapping Methods 0.000 description 4
- 238000013461 design Methods 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 3
- 238000013528 artificial neural network Methods 0.000 description 2
- 238000010276 construction Methods 0.000 description 2
- 230000008569 process Effects 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 230000007306 turnover Effects 0.000 description 2
- 230000003044 adaptive effect Effects 0.000 description 1
- 230000008901 benefit Effects 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 230000007812 deficiency Effects 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000009792 diffusion process Methods 0.000 description 1
- 235000013399 edible fruits Nutrition 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 210000000744 eyelid Anatomy 0.000 description 1
- 238000003384 imaging method Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000005764 inhibitory process Effects 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000001537 neural effect Effects 0.000 description 1
- 238000007781 pre-processing Methods 0.000 description 1
- 238000002360 preparation method Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000004088 simulation Methods 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
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20084—Artificial neural networks [ANN]
Landscapes
- Engineering & Computer Science (AREA)
- 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 discloses a kind of CNN denoising method extracted based on Concurrent Feature, including six steps: step 1, builds the CNN denoising network model of Concurrent Feature extraction；Step 2, the training parameter of initialization CNN denoising network model；Step 3: building training set；Step 4, allowable loss function, and train CNN to denoise network model to minimize loss function as target, obtain CNN denoising model；Step 5, using noise image as the input of CNN denoising model, output is the noise information that network model learns；Step 6 subtracts the noise information that step 5 learns with noise image, the clean image after denoising can be obtained.The present invention can removal noise more thoroughly, the texture information of image can be effectively maintained, and significantly improve objective indicator PSNR and SSIM.
Description
Technical field
The present invention relates to image denoising field more particularly to a kind of CNN denoising methods extracted based on Concurrent Feature.
Background technique
Digital picture in reality is subjected to imaging device in digitlization and transmission process and external environmental noise is interfered
Deng influence, referred to as noisy image or noise image.The final purpose of image denoising is to improve given image, solves real image
The problem of leading to image quality decrease due to noise jamming.Picture quality can be effectively improved by noise-removed technology, is increased
Signal-to-noise ratio preferably embodies information entrained by original image, and as a kind of important preprocessing means, people are to image denoising
Algorithm conducts extensive research.
The classical way of image denoising has much at present, but can be roughly divided into two types, and one kind is to be based on filter in spatial domain,
Such as mean filter, median filtering；Another kind of filtered based on transform domain, most such as the Bayes in Gauss scale mixed model
Small square law.Existing Denoising Algorithm, some obtain preferable effect in low-dimensional signal and image processing, are not suitable for higher-dimension but
Signal and image processing；Or denoising effect is preferable, but lost part image edge information；Or it is dedicated to studying detection image side
Edge information retains image detail, does not filter in global scope but, also do not account for contacting between natural image block and block
Property.Therefore the denoising effect that existing method integrally obtains is unsatisfactory.
In order to solve the problems, such as that traditional denoising method exists, neural network is used for image denoising.A kind of patent " CNN-
The intelligent filter method and system of LMS picture noise " (patent No.: 201810128238.0), disclose it is a kind of by LMS from
It is embedded in CNN intelligent control model in adaptive filtering system, adjusts LMS Avaptive filtering system parameter, picture noise is filtered
Wave or inhibition, to remove the method that picture noise obtains filtering image.A kind of patent " figure based on compression-type convolutional neural networks
As denoising method " (patent No.: 201710286383.7), disclose a kind of convolutional layer by original denoising convolutional neural networks
It has been substituted for and has decomposed compressed convolutional layer, and the method for carrying out image denoising via low-rank matrix." one kind is based on ReLU to patent
The image de-noising method of convolutional neural networks " (patent No.: 201610482594.3), discloses a kind of based on ReLU convolutional Neural
The real-time de-noising method of network model.They and the present invention design the difference is that:
(1) the parallel MPFE characteristic extracting module of two-way is devised in the present invention.
(2) present invention using intensive connection by before feature that bottom extracts to being transmitted to higher.
(3) Fusion Features for the different scale that the feature and MPFE characteristic extracting module that the present invention extracts bottom shift to an earlier date,
To make the feature extracted characterize image information to the greatest extent.
The present invention and " a kind of intelligent filter method and system of CNN-LMS picture noise ", " one kind is based on ReLU convolution mind
Image de-noising method through network " is compared, advantage with " a kind of image de-noising method based on ReLU convolutional neural networks " are as follows:
(1) present invention has carried out zero padding operation before each convolutional layer, guarantees the size for not changing image, and can use up can
The marginal information of the reservation image of energy.
(2) present invention uses 5 characteristic extracting module MPFE, it is the parallel network of a two-way, respectively using not
Same convolution kernel extracts different features, is easy to implement high quality denoising.
(3) present invention is linked together the characteristics of image that different depth proposes using intensive connection, by Fusion Features
With it is rear, guarantee network make full use of the various feature learnings of image output and input between mapping relations.
The purpose of the present invention is to provide a kind of image de-noising methods of high quality, retain as far as possible while denoising
The marginal information and detailed information of image.
Summary of the invention
To solve the deficiencies in the prior art, the purpose of the present invention is to provide a kind of images based on multiple dimensioned parallel C NN
Denoising method, to improve the denoising effect of image.
The present invention relates to a kind of CNN denoising methods extracted based on Concurrent Feature, which is characterized in that specifically according to following step
It is rapid to carry out:
Step 1 builds the CNN denoising network model of Concurrent Feature extraction；
Step 2, the training parameter of initialization CNN denoising network model；
Step 3: building training set；
Step 4, allowable loss function, and train CNN to denoise network model to minimize loss function as target, it obtains
To CNN denoising model；
Step 5, using noise image as the input of CNN denoising model, output is the noise that network model learns
Information；
Step 6 subtracts the noise information that step 5 learns with noise image, the clean image after denoising can be obtained.
Further, a kind of CNN denoising method extracted based on Concurrent Feature, which is characterized in that the step
In one, the CNN denoising network model that the Concurrent Feature built is extracted includes 5 characteristic extracting module MPFE, it is a two-way
Parallel network is the convolution kernel of 3 × 3 series connection 5 × 5 on one side, and another side is the convolution kernel of 5 × 5 series connection 3 × 3, finally by two-way
Fusion Features are carried out, the mathematical model of MPFE is,
Wherein n=1,2 ..., 5, ω and b respectively represent weight and biasing, and subscript indicates that the number of plies at place, subscript represent
Convolution kernel size, d indicate input channel, MPnIAnd MPnOIndicate outputting and inputting for n-th of MPFE, [MPnI,A2,B2] indicate special
The serial operation of sign.
Further, a kind of CNN denoising method extracted based on Concurrent Feature, which is characterized in that described to build
Concurrent Feature extract CNN denoising network model include 5 characteristic extracting module MPFE in, first and second MPFE
The mathematical model of input be,
Wherein target meaning is identical as in formula (1) up and down.
Further, a kind of CNN denoising method extracted based on Concurrent Feature, which is characterized in that the step
In one, the CNN denoising network model that the Concurrent Feature built is extracted includes 22 convolutional layers, and the size of convolution kernel is 3 × 3 or 1
× 1, wherein having an activation primitive behind 3 × 3 convolution kernels is the active coating of ReLU, CNN denoises the mathematical modulo of network model
Type is,
Wherein target meaning is identical as in formula (1) up and down.
Further, a kind of CNN denoising method extracted based on Concurrent Feature, which is characterized in that the step
In two, the training parameter of CNN denoising network model is specifically configured to: being trained for 75 generations altogether, is used Adam as optimizer, study effect
The initial value of rate is set as 0.001, is set as 64, steps_ every the batch_size in 10 generation drop by half, every generation
Per_epoch is set as 2000.
Further, a kind of CNN denoising method extracted based on Concurrent Feature, which is characterized in that the step
In three, the construction method of training set is to be added at random to 400 180 × 180 normal grayscale images and determine that the Gauss of concentration makes an uproar
After sound；Standard picture is cut into multiple 40 × 40 image blocks according to step-length 10；Again each image block turn over up and down
Turn, the operation such as rotation at any angle, finally obtains 23.84 ten thousand width image blocks, form training set.
Further, a kind of CNN denoising method extracted based on Concurrent Feature, which is characterized in that the step
In four, the loss function of design is,
Wherein YiWithThe clean image for respectively indicating the corresponding ideal clean image of the i-th amplitude and noise acoustic image and estimating, ω
Weight and biasing are respectively represented with b.
The present invention achieves following technical effect compared with the existing technology:
(1) present invention has carried out zero padding operation before each convolutional layer, guarantees the size for not changing image, and can use up can
The marginal information of the reservation image of energy.
(2) present invention uses 5 characteristic extracting module MPFE, it is the parallel network of a two-way, respectively using not
Same convolution kernel extracts different features, is easy to implement high quality denoising.
(3) present invention is linked together the characteristics of image that different depth proposes using intensive connection, by Fusion Features
With it is rear, guarantee network make full use of the various feature learnings of image output and input between mapping relations.
Detailed description of the invention
It in order to more clearly explain the embodiment of the invention or the technical proposal in the existing technology, below will be to institute in embodiment
Attached drawing to be used is needed to be briefly described, it should be apparent that, the accompanying drawings in the following description is only some implementations of the invention
Example, for those of ordinary skill in the art, without any creative labor, can also be according to these attached drawings
Obtain other attached drawings.
Fig. 1 is denoising flow chart of the invention；
Fig. 2 is the structure chart of characteristic extracting module MPFE；
Fig. 3 is that the CNN extracted based on Concurrent Feature denoises network model figure；
Fig. 4 is 6 kinds of widely used test images；
Fig. 5 is the denoising result figure of the present invention and existing denoising method；
Wherein (a) standard picture, (b) noise image/14.14dB, (c) result/29.85dB of BM3D, (d) knot of WNNM
Fruit/30.28dB, (e) result/29.08dB of EPLL, (f) result/29.53dB of TNRD, (g) result/29.94dB of MLP,
(h) result/30.36dB of DnCNN-S, (i) result/30.59dB of the invention.
Specific embodiment
Following will be combined with the drawings in the embodiments of the present invention, and technical solution in the embodiment of the present invention carries out clear, complete
Site preparation description, it is clear that described embodiments are only a part of the embodiments of the present invention, instead of all the embodiments.It is based on
Embodiment in the present invention, it is obtained by those of ordinary skill in the art without making creative efforts every other
Embodiment shall fall within the protection scope of the present invention.
As shown in Fig. 2, the present invention discloses a kind of CNN denoising method extracted based on Concurrent Feature, including six steps.Step
S1 builds the CNN denoising network model of Concurrent Feature extraction；Step S2, the training parameter of initialization CNN denoising network model；
Step S3 constructs training set；Step S4, allowable loss function, and train CNN to denoise net to minimize loss function as target
Network model obtains CNN denoising model；Step S5, using noise image as the input of CNN denoising model, output is network
The noise information that model learning arrives；Step S6 subtracts the noise information that step 5 learns with noise image, denoising can be obtained
Clean image afterwards.
Further, a kind of CNN denoising method extracted based on Concurrent Feature, which is characterized in that the step
In one, the CNN denoising network model that the Concurrent Feature built is extracted includes 5 characteristic extracting module MPFE, such as Fig. 2, it is one
The parallel network of a two-way is the convolution kernel of 3 × 3 series connection 5 × 5 on one side, and another side is the convolution kernel of 5 × 5 series connection 3 × 3, finally
Two-way is subjected to Fusion Features, the mathematical model of MPFE is,
Wherein n=1,2 ..., 5, ω and b respectively represent weight and biasing, and subscript indicates that the number of plies at place, subscript represent
Convolution kernel size, d indicate input channel, MPnIAnd MPnOIndicate outputting and inputting for n-th of MPFE, [MPnI,A2,B2] indicate special
The serial operation of sign.
Further, a kind of CNN denoising method extracted based on Concurrent Feature, which is characterized in that described to build
Concurrent Feature extract CNN denoising network model include 5 characteristic extracting module MPFE in, first and second MPFE
The mathematical model of input be,
Wherein target meaning is identical as in formula (1) up and down.
Further, a kind of CNN denoising method extracted based on Concurrent Feature, which is characterized in that the step
In one, such as Fig. 3, the CNN denoising network model that the Concurrent Feature built is extracted includes 22 convolutional layers, and the size of convolution kernel is 3
× 3 or 1 × 1, wherein having an activation primitive behind 3 × 3 convolution kernels is the active coating of ReLU, CNN denoises network model
Mathematical model is,
Wherein target meaning is identical as in formula (1) up and down.
Further, a kind of CNN denoising method extracted based on Concurrent Feature, which is characterized in that the step
In two, the training parameter of CNN denoising network model is specifically configured to: being trained for 75 generations altogether, is used Adam as optimizer, study effect
The initial value of rate is set as 0.001, is set as 64, steps_ every the batch_size in 10 generation drop by half, every generation
Per_epoch is set as 2000.
Further, a kind of CNN denoising method extracted based on Concurrent Feature, which is characterized in that the step
In three, the construction method of training set is to be added at random to 400 180 × 180 normal grayscale images and determine that the Gauss of concentration makes an uproar
After sound；Standard picture is cut into multiple 40 × 40 image blocks according to step-length 10；Again each image block turn over up and down
Turn, the operation such as rotation at any angle, finally obtains 23.84 ten thousand width image blocks, form training set.
Further, a kind of CNN denoising method extracted based on Concurrent Feature, which is characterized in that the step
In four, the loss function of design is,
Wherein YiWithThe clean image for respectively indicating the corresponding ideal clean image of the i-th amplitude and noise acoustic image and estimating, ω
Weight and biasing are respectively represented with b.
With formula (4) for objective function, CNN denoising model of the invention is trained using training set, available input
Mapping function between noise image and output(it is exactly the noise information that CNN denoising model learns), then use
Input noise image subtracts mapping functionClean image after denoising can be obtained.
In order to verify effectiveness of the invention, l-G simulation test has been carried out.
Experiment is on the PC of Intel (R) Core (TM) i5-8300H CPU2.30GHz and Nvidia1050Ti GPU
It is run in Keras environment.
Training parameter is specifically configured to: being trained for 75 generations altogether, is used Adam as optimizer, the initial value of learning efficiency is arranged
It is 0.001, is set as 64, steps_per_epoch every the batch_size in 10 generation drop by half, every generation and is set as
2000.With 23.84 ten thousand 40 × 40 image blocks, training set is formed.Utilize this training set training denoising model of the invention, root
According to the difference of test set, two experiments have been carried out, and have been compared respectively with several advanced denoising methods.The method compared includes:
BM3D(K.Dabov,et al,Image denoising by sparse 3-D transform-domain
Collaborative filtering, IEEE Trans.Image Process., 2007,16 (8): 2080-2095), WNNM
(S.Gu,et al,Weighted nuclear norm minimization with application to image
Denoising, in Proc.IEEE Conf.Comput.Vis.Pattern Recognit., 2014:2862-2869), CSF
(U.Schmidt et al.,Shrinkage fields for effective image restoration,in
Proc.IEEE Conf.Comput.Vis.Pattern Recognit.,pp.2774-2781,Columbus,OH,USA
(2014)),EPLL(D.Zoran,et al,From learning models ofnatural image patches to
Whole image restoration, in Proc.IEEE Int.Conf.Comput.Vis., 2011:479-486), TNRD
(Y.Chen,et al,Trainable nonlinear reaction diffusion:A flexible framework for
fast and effective image restoration,IEEE Trans.Pattern Anal.Mach.Intell.,
2017,39(6):1256-1272),MLP(H.C.Burger,et al,Image denoising:Canplain neural
Networks compete with BM3D?, in Proc.IEEE Conf.Comput.Vis.Pattern Recognit.,
2012:2392-2399) and DnCNN-S (K.Zhang et al., Beyond a gaussian denoiser:Residual
learning of deep CNN for image denoising,IEEE Trans.Image Process.26(7),3142-
3155(2017)).Using objectively evaluating index peak signal-to-noise ratio (Peak Signal to Noise Ratio, PSNR) and structure
Similitude (Structural similarity, SSIM) measures denoising effect.In general, the value of PSNR is bigger, and noise is to letter
Number annoyance level is weaker, and SSIM is bigger, and image fault is fewer, illustrates that the denoising effect of image is better.
Experiment one, using the image in Fig. 4 as test image, table 1 is test result, and black runic indicates that highest refers to
Mark.In an experiment, noise level σ is respectively set to 15,25,50, from test result as can be seen that objective indicator of the invention is equal
Higher than other methods, it is better than the effect of other several denoising methods to denoise effect.
The method of the invention of table 1 and several advanced method comparison results
Experiment two, denoising effect in order to further illustrate the present invention has chosen BSD68 as test set, and at first
Into several method be compared, the results are shown in Table 2:
Test (PSNR) result of table 2 on BSD68 test set
Can be seen that denoising method of the invention from 2 test result of table can obtain better PSNR and SSIM.
Fig. 5 is the result figure after distinct methods and the method for the present invention denoise the image that 50% Gaussian noise is added
Picture, it is clear that the present invention protects the marginal information of image well, remains the characteristic information of image, such as double-edged eyelid feature,
Obtain best denoising effect.
The foregoing is merely one embodiment of the present of invention, are not intended to limit the scope of the present invention.It is all
Any modification, equivalent replacement, improvement and so within the spirit and principles in the present invention, are all contained in protection scope of the present invention
It is interior.
Claims (5)
1. a kind of CNN denoising method extracted based on Concurrent Feature, which is characterized in that specifically follow the steps below:
Step 1 builds the CNN denoising network model of Concurrent Feature extraction；
Step 2, the training parameter of initialization CNN denoising network model；
Step 3: building training set；
Step 4, allowable loss function, and train CNN to denoise network model to minimize loss function as target, obtain CNN
Denoising model；
Step 5, using noise image as the input of CNN denoising model, output is the noise letter that network model learns
Breath；
Step 6 subtracts the noise information that step 5 learns with noise image, the clean image after denoising can be obtained.
2. a kind of CNN denoising method extracted based on Concurrent Feature according to claim 1, which is characterized in that the step
In rapid one, the CNN denoising network model that the Concurrent Feature built is extracted includes 5 characteristic extracting module MPFE, it is one two
The parallel network in road is the convolution kernel of 3 × 3 series connection 5 × 5 on one side, and another side is the convolution kernel of 5 × 5 series connection 3 × 3, finally by two
Road carries out Fusion Features, and the mathematical model of MPFE is,
Wherein n=1,2 ..., 5, ω and b respectively represent weight and biasing, and subscript indicates that the number of plies at place, subscript represent convolution
Core size, d indicate input channel, MPnIAnd MPnOIndicate outputting and inputting for n-th of MPFE, [MPnI,A2,B2] indicate feature
Serial operation.
3. a kind of CNN denoising method extracted based on Concurrent Feature according to claim 2, which is characterized in that described to take
In 5 characteristic extracting module MPFE that the CNN denoising network model that the Concurrent Feature built is extracted includes, first and second
The mathematical model of the input of MPFE is,
Wherein target meaning is identical as in formula (1) up and down.
4. a kind of CNN denoising method extracted based on Concurrent Feature according to claim 1, which is characterized in that the step
In rapid one, the CNN denoising network model that the Concurrent Feature built is extracted includes 22 convolutional layers, the size of convolution kernel be 3 × 3 or
1 × 1, wherein having an activation primitive behind 3 × 3 convolution kernels is the active coating of ReLU, CNN denoises the mathematical modulo of network model
Type is,
Wherein target meaning is identical as in formula (1) up and down.
5. a kind of CNN denoising method extracted based on Concurrent Feature according to claim 1, which is characterized in that the step
In rapid two, the training parameter of CNN denoising network model is specifically configured to: being trained for 75 generations altogether, is used Adam as optimizer, study
The initial value of efficiency is set as 0.001, is set as 64, steps_ every the batch_size in 10 generation drop by half, every generation
Per_epoch is set as 2000.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811323811.XA CN109410149B (en) | 2018-11-08 | 2018-11-08 | CNN denoising method based on parallel feature extraction |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811323811.XA CN109410149B (en) | 2018-11-08 | 2018-11-08 | CNN denoising method based on parallel feature extraction |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109410149A true CN109410149A (en) | 2019-03-01 |
| CN109410149B CN109410149B (en) | 2019-12-31 |
Family
ID=65472193
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811323811.XA Active CN109410149B (en) | 2018-11-08 | 2018-11-08 | CNN denoising method based on parallel feature extraction |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109410149B (en) |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110189260A (en) * | 2019-04-15 | 2019-08-30 | 浙江大学 | A kind of image denoising method based on multiple dimensioned parallel gate neural network |
| CN110782414A (en) * | 2019-10-30 | 2020-02-11 | 福州大学 | Dark light image denoising method based on dense connection convolution |
| CN110852966A (en) * | 2019-11-04 | 2020-02-28 | 西北工业大学 | Image noise estimation method based on deep convolutional neural network |
| CN111275643A (en) * | 2020-01-20 | 2020-06-12 | 西南科技大学 | True noise blind denoising network model and method based on channel and space attention |
| CN111667424A (en) * | 2020-05-28 | 2020-09-15 | 武汉大学 | Unsupervised real image denoising method |
| CN111860273A (en) * | 2020-07-14 | 2020-10-30 | 吉林大学 | Magnetic resonance underground water detection noise suppression method based on convolutional neural network |
| CN112488934A (en) * | 2020-11-26 | 2021-03-12 | 杭州电子科技大学 | Finger vein image denoising method based on CS-TCGAN |
Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106875361A (en) * | 2017-02-17 | 2017-06-20 | 深圳市唯特视科技有限公司 | A kind of method that poisson noise is removed based on depth convolutional neural networks |
| CN107516304A (en) * | 2017-09-07 | 2017-12-26 | 广东工业大学 | A kind of image de-noising method and device |
| CN107590518A (en) * | 2017-08-14 | 2018-01-16 | 华南理工大学 | A kind of confrontation network training method of multiple features study |
| 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 |
| CN107862705A (en) * | 2017-11-21 | 2018-03-30 | 重庆邮电大学 | A kind of unmanned plane small target detecting method based on motion feature and deep learning feature |
| CN108280811A (en) * | 2018-01-23 | 2018-07-13 | 哈尔滨工业大学深圳研究生院 | A kind of image de-noising method and system based on neural network |
| US10032256B1 (en) * | 2016-11-18 | 2018-07-24 | The Florida State University Research Foundation, Inc. | System and method for image processing using automatically estimated tuning parameters |
| CN108446661A (en) * | 2018-04-01 | 2018-08-24 | 桂林电子科技大学 | A kind of deep learning parallelization face identification method |
| CN108564555A (en) * | 2018-05-11 | 2018-09-21 | 中北大学 | A kind of digital image noise reduction method based on NSST and CNN |
| US20180293711A1 (en) * | 2017-04-06 | 2018-10-11 | Disney Enterprises, Inc. | Kernel-predicting convolutional neural networks for denoising |
-
2018
- 2018-11-08 CN CN201811323811.XA patent/CN109410149B/en active Active
Patent Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US10032256B1 (en) * | 2016-11-18 | 2018-07-24 | The Florida State University Research Foundation, Inc. | System and method for image processing using automatically estimated tuning parameters |
| CN106875361A (en) * | 2017-02-17 | 2017-06-20 | 深圳市唯特视科技有限公司 | A kind of method that poisson noise is removed based on depth convolutional neural networks |
| US20180293711A1 (en) * | 2017-04-06 | 2018-10-11 | Disney Enterprises, Inc. | Kernel-predicting convolutional neural networks for denoising |
| CN107590518A (en) * | 2017-08-14 | 2018-01-16 | 华南理工大学 | A kind of confrontation network training method of multiple features study |
| 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 |
| CN107516304A (en) * | 2017-09-07 | 2017-12-26 | 广东工业大学 | A kind of image de-noising method and device |
| CN107862705A (en) * | 2017-11-21 | 2018-03-30 | 重庆邮电大学 | A kind of unmanned plane small target detecting method based on motion feature and deep learning feature |
| CN108280811A (en) * | 2018-01-23 | 2018-07-13 | 哈尔滨工业大学深圳研究生院 | A kind of image de-noising method and system based on neural network |
| CN108446661A (en) * | 2018-04-01 | 2018-08-24 | 桂林电子科技大学 | A kind of deep learning parallelization face identification method |
| CN108564555A (en) * | 2018-05-11 | 2018-09-21 | 中北大学 | A kind of digital image noise reduction method based on NSST and CNN |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110189260A (en) * | 2019-04-15 | 2019-08-30 | 浙江大学 | A kind of image denoising method based on multiple dimensioned parallel gate neural network |
| CN110782414A (en) * | 2019-10-30 | 2020-02-11 | 福州大学 | Dark light image denoising method based on dense connection convolution |
| CN110852966A (en) * | 2019-11-04 | 2020-02-28 | 西北工业大学 | Image noise estimation method based on deep convolutional neural network |
| CN111275643A (en) * | 2020-01-20 | 2020-06-12 | 西南科技大学 | True noise blind denoising network model and method based on channel and space attention |
| CN111667424A (en) * | 2020-05-28 | 2020-09-15 | 武汉大学 | Unsupervised real image denoising method |
| CN111667424B (en) * | 2020-05-28 | 2022-04-01 | 武汉大学 | Unsupervised real image denoising method |
| CN111860273A (en) * | 2020-07-14 | 2020-10-30 | 吉林大学 | Magnetic resonance underground water detection noise suppression method based on convolutional neural network |
| CN111860273B (en) * | 2020-07-14 | 2022-07-05 | 吉林大学 | Magnetic resonance underground water detection noise suppression method based on convolutional neural network |
| CN112488934A (en) * | 2020-11-26 | 2021-03-12 | 杭州电子科技大学 | Finger vein image denoising method based on CS-TCGAN |
| CN112488934B (en) * | 2020-11-26 | 2024-02-09 | 杭州电子科技大学 | CS-TCGAN-based finger vein image denoising method |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109410149B (en) | 2019-12-31 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109410149A (en) | A kind of CNN denoising method extracted based on Concurrent Feature | |
| CN109003240A (en) | A kind of image de-noising method based on multiple dimensioned parallel C NN | |
| CN106204468B (en) | A kind of image de-noising method based on ReLU convolutional neural networks | |
| CN110120020A (en) | A kind of SAR image denoising method based on multiple dimensioned empty residual error attention network | |
| CN109118435A (en) | A kind of depth residual error convolutional neural networks image de-noising method based on PReLU | |
| CN111723701B (en) | Underwater target identification method | |
| CN108416755A (en) | A kind of image de-noising method and system based on deep learning | |
| CN101303764B (en) | Method for self-adaption amalgamation of multi-sensor image based on non-lower sampling profile wave | |
| CN109035142A (en) | A kind of satellite image ultra-resolution method fighting network integration Aerial Images priori | |
| CN109214990A (en) | A kind of depth convolutional neural networks image de-noising method based on Inception model | |
| CN106328150A (en) | Bowel sound detection method, device and system under noisy environment | |
| CN105046277A (en) | Robust mechanism research method of characteristic significance in image quality evaluation | |
| CN107292336A (en) | A kind of Classification of Polarimetric SAR Image method based on DCGAN | |
| CN106600595A (en) | Human body characteristic dimension automatic measuring method based on artificial intelligence algorithm | |
| CN110503613A (en) | Based on the empty convolutional neural networks of cascade towards removing rain based on single image method | |
| CN109859120A (en) | Image defogging method based on multiple dimensioned residual error network | |
| CN101685158B (en) | Hidden Markov tree model based method for de-noising SAR image | |
| CN110456332A (en) | A kind of underwater sound signal Enhancement Method based on autocoder | |
| CN113723171B (en) | Electroencephalogram signal denoising method based on residual error generation countermeasure network | |
| CN108230280A (en) | Image speckle noise minimizing technology based on tensor model and compressive sensing theory | |
| CN106504207A (en) | A kind of image processing method | |
| Yang et al. | An adaptive contourlet HMM–PCNN model of sparse representation for image denoising | |
| CN115442191A (en) | Communication signal noise reduction method and system based on relative average generation countermeasure network | |
| CN104853182A (en) | Amplitude and phase based stereo image quality objective evaluation method | |
| CN112785539A (en) | Multi-focus image fusion method based on image adaptive decomposition and parameter adaptive |
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 |