CN109671026A - Gray level image noise-reduction method based on empty convolution and automatic encoding and decoding neural network - Google Patents
Gray level image noise-reduction method based on empty convolution and automatic encoding and decoding neural network Download PDFInfo
- Publication number
- CN109671026A CN109671026A CN201811434831.4A CN201811434831A CN109671026A CN 109671026 A CN109671026 A CN 109671026A CN 201811434831 A CN201811434831 A CN 201811434831A CN 109671026 A CN109671026 A CN 109671026A
- Authority
- CN
- China
- Prior art keywords
- image
- noise
- convolution
- gray level
- neural network
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 36
- 238000013528 artificial neural network Methods 0.000 title claims abstract description 19
- 230000009467 reduction Effects 0.000 claims abstract description 41
- 238000012549 training Methods 0.000 claims abstract description 34
- 238000010606 normalization Methods 0.000 claims description 15
- 239000011248 coating agent Substances 0.000 claims description 12
- 238000000576 coating method Methods 0.000 claims description 12
- 238000003475 lamination Methods 0.000 claims description 8
- 238000000605 extraction Methods 0.000 claims description 7
- 238000012545 processing Methods 0.000 claims description 6
- 230000008569 process Effects 0.000 claims description 5
- 230000004913 activation Effects 0.000 claims description 2
- 230000000007 visual effect Effects 0.000 abstract description 4
- 238000001914 filtration Methods 0.000 description 7
- 230000008901 benefit Effects 0.000 description 3
- 238000011161 development Methods 0.000 description 3
- 238000010586 diagram Methods 0.000 description 3
- 230000006870 function Effects 0.000 description 3
- 230000009286 beneficial effect Effects 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000004519 manufacturing process Methods 0.000 description 2
- 210000004218 nerve net Anatomy 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 230000006835 compression Effects 0.000 description 1
- 238000007906 compression Methods 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000007796 conventional method Methods 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000004927 fusion Effects 0.000 description 1
- 238000003706 image smoothing Methods 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000001537 neural effect Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000000644 propagated effect Effects 0.000 description 1
- 238000011084 recovery Methods 0.000 description 1
- 230000003362 replicative effect Effects 0.000 description 1
- 238000012546 transfer Methods 0.000 description 1
- 230000009466 transformation 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
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- 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)
- Compression Or Coding Systems Of Tv Signals (AREA)
- Image Processing (AREA)
Abstract
The invention discloses a kind of gray level image noise-reduction method based on empty convolution and automatic encoding and decoding neural network, comprising the following steps: using clear image and noise image corresponding with clear image as a training sample, training set is constructed with this；Construct image noise reduction model, the image noise reduction model includes the characteristics of image decoding unit that characteristics of image coding unit, the coding characteristic figure for exporting to characteristics of image coding unit for carrying out feature coding to noise image are decoded, wherein, characteristics of image coding unit includes characteristic extracting module and 10 coding modules, and image decoding unit includes 10 decoder modules and image restoration module；Using the image noise reduction model of training set training building, trained image noise reduction model is obtained；In application, noise image is input to trained image noise reduction model, it is computed output noise-reduced image.The gray level image noise-reduction method can rapidly remove the noise of image, promote the visual effect of image.
Description
Technical field
The invention belongs to image signal process fields, and in particular to one kind is based on empty convolution and automatic encoding and decoding nerve net
The gray level image noise-reduction method of network.
Background technique
Image is that people obtain the extremely important source of information.It is universal with digital equipment in the information age of today,
Digital picture has become the important means that people obtain information, is deep into the every aspect of production and life, achieves huge
Social and economic benefit.The combination of the research fields such as image processing techniques and machine learning, machine vision in recent years, even more generates
Unprecedented new development and breakthrough.However during image acquisition, processing, compression, transmitting, storing and replicating, no
It can avoid that noise can be introduced, to reduce picture quality.The main target of image denoising is to filter out random noise therein, and use up
Possibly retain image detail information.With the development of science and technology, requirement of each process to picture quality be not yet in image procossing
It is disconnected to improve.Therefore, research Image denoising algorithm is very necessary and important.
Image denoising is related to the numerous areas such as optical system, microelectric technique, computer science, mathematical analysis, is one
Comprehensive extremely strong basic subject, and have extremely important status in field of image processing.Researcher proposes in large quantities
This process is known as image smoothing or image filtering to eliminate the noise in figure by Denoising Algorithm.According to the filtering side of use
Method, image filtering can be divided into two classes: one kind being referred to as linear filtering, is a kind of simplest removal picture noise mode；One
Kind is denoised using original signal and the distinctive statistical property of noise signal, is referred to as nonlinear filtering.Meanwhile according to filter
Signal domain existing for wave, image denoising can be divided into two classes: one kind is Space domain, mainly in image space domain to pixel
Point is handled；Another kind of is transform domain method, is modified processing to image coefficient in transform domain, then passes through inverse transformation
Spatial domain picture after obtaining final process.
With the high speed development of modern processors computing capability and deep learning theory, image noise reduction neural network based
Method is as a kind of novel image de-noising method, it has also become hot spot.With conventional filter (Gaussian filter, median filter)
Denoising is compared, and image noise reduction neural network based has many advantages, such as that image is apparent, the time is shorter.Simultaneously as nerve net
The expansibility of network model itself is extremely strong, and denoising method neural network based can include the excellent of substantially all conventional methods
Point: it both can be in conjunction with airspace and transform domain, while also embodying linear filtering and nonlinear filtering.Although deep neural network
Method can obtain better effect, but such methods still need to occupy biggish memory and great computing resource, it is difficult in reality
It is used in border life production process.Therefore, how more efficient more compactly by constructing the side for training neural fusion to denoise
Method has important engineering practical value and theory directive significance.
Summary of the invention
The object of the present invention is to provide a kind of gray level image noise reduction based on empty convolution and automatic encoding and decoding neural network
Method, this method can quickly remove the noise of image, promote the visual effect of image.
For achieving the above object, the present invention the following technical schemes are provided:
A kind of gray level image noise-reduction method based on empty convolution and automatic encoding and decoding neural network, comprising the following steps:
Using clear image and noise image corresponding with clear image as a training sample, training set is constructed with this；
Image noise reduction model is constructed, which includes special for carrying out the image of feature coding to noise image
The characteristics of image decoding unit that sign coding unit, the coding characteristic figure for exporting to characteristics of image coding unit are decoded,
Wherein, characteristics of image coding unit includes characteristic extracting module and 10 coding modules, and image decoding unit includes 10 decodings
Module and image restoration module；
Using the image noise reduction model of training set training building, trained image noise reduction model is obtained；
In application, noise image is input to trained image noise reduction model, it is computed output noise-reduced image.
The gray level image noise-reduction method has the beneficial effect that
The structure of image noise reduction model is very succinct in the present invention, and after training, the byte that parameter occupies is seldom, only
0.5M is suitable for some non-embedded type mini systems, furthermore, which can rapidly remove making an uproar for image
Sound promotes the visual effect of image.
Detailed description of the invention
In order to more clearly explain the embodiment of the invention or the technical proposal in the existing technology, to embodiment or will show below
There is attached drawing needed in technical description to do simply to introduce, it should be apparent that, the accompanying drawings in the following description is only this
Some embodiments of invention for those of ordinary skill in the art, can be with root under the premise of not making the creative labor
Other accompanying drawings are obtained according to these attached drawings.
Fig. 1 is the structural schematic diagram for the image noise reduction model that embodiment provides；
Fig. 2 is the structural schematic diagram of coding module in Fig. 1；
Fig. 3 is the structural schematic diagram of decoder module in Fig. 2；
Fig. 4 is the flow chart of the building that embodiment provides and training image noise reduction model；
Fig. 5 is the flow chart that image noise reduction is carried out using image noise reduction model that embodiment provides.
Specific embodiment
To make the objectives, technical solutions, and advantages of the present invention more comprehensible, with reference to the accompanying drawings and embodiments to this
Invention is described in further detail.It should be appreciated that the specific embodiments described herein are only used to explain the present invention,
And the scope of protection of the present invention is not limited.
In order to realize to image denoising, the present embodiment provides a kind of based on empty convolution and automatic encoding and decoding neural network
Gray level image noise-reduction method specifically includes image noise reduction model construction and using the image noise reduction model to noise image denoising two
A part.
Image noise reduction model is constructed, as shown in figure 4, including following procedure:
Prepare training set first, i.e., Gaussian noise added to clear image using fixed noise grade, obtain with clearly
The corresponding noise image of image, and using clear image and noise image corresponding with clear image as a training sample, with
This building training set.
In order to adapt to the input image size of image noise reduction model, using identical arbitrary width by clear image and with it is clear
The corresponding noise image of image is divided into several groups image block, using every group of image block as a training sample.It specifically, can be with
Clear image and noise image corresponding with clear image are divided into 40 × 40 image block.
Then image noise reduction model is built, which is made of empty convolution, automatic encoding and decoding neural network, such as Fig. 1 institute
Show, specifically includes the characteristics of image coding unit for carrying out feature coding to noise image and for encoding list to characteristics of image
The characteristics of image decoding unit that the coding characteristic figure of member output is decoded.
Wherein, characteristics of image coding unit include characteristic extracting module F1 and coding module E1, coding module E2 ..., compile
Code module E10, specifically, as shown in Fig. 2, it is that 3 × 3, convolution kernel number is that each characteristic extracting module, which includes convolution kernel size,
64 convolutional layer and active coating；Coding module include sequentially connected dimensionality reduction submodule, feature extraction submodule and liter dimension
Submodule, wherein it is the convolution that 1 × 1, convolution kernel number is 32 that dimensionality reduction submodule, which includes a convolution kernel with liter dimension submodule,
Layer CONV, the normalization layer BN of connection convolutional layer output, the active coating RELU of connection normalization layer output；Feature extraction submodule
It is that 1 × 1, convolution kernel number is 32, uses expansion rate for the convolutional layer CONV of 2 empty convolution, connection volume including a convolution kernel
The normalization layer BN of lamination output, the active coating RELU of connection normalization layer output.
Characteristics of image decoding unit include decoder module D1, decoder module D2 ..., decoder module D10 and image restoration
Module R1.Specifically, as shown in figure 3, each decoder module include sequentially connected dimensionality reduction submodule, feature extraction submodule with
And liter tie up submodule, wherein dimensionality reduction submodule and rise dimension submodule include a convolution kernel be 1 × 1, convolution kernel number be
32 warp lamination DECONV, the normalization layer BN of connection warp lamination output, the active coating of connection normalization layer output；Feature
Extracting sub-module includes that a convolution kernel is that 1 × 1, convolution kernel number is 32, uses expansion rate for the warp lamination of 2 empty convolution
DECONV, the normalization layer BN of connection warp lamination output, the active coating RELU of connection normalization layer output；Image restoration module
It is the convolutional layer that 3 × 3, convolution kernel number is 64 including a convolution kernel size.
Specifically, the activation primitive of active coating is all made of ReLU function.
When training, using the image noise reduction model of training set training building, trained image noise reduction model is obtained, that is, is instructed
During white silk, distorted image successively after characteristics of image coding unit and the processing of characteristics of image decoding unit, is schemed according to clear
Loss as calculating each costing bio disturbance unit, and superposition is weighted to each loss and is finally lost, then using most
Loss backpropagation eventually updates network weight parameter.
In the present embodiment, training set is 400 gray scale pictures for being derived from BSD500 data set, i.e., to 400 pictures according to
10 fixed intervals step-length is cut into 40 × 40 image block, available 238400 image blocks is amounted to, as training set.
In training image noise reduction model, input layer size is 40 × 40；When carrying out image restoration operation, input layer is big
The small actual size for image.For the loss function used when training for L2 loss function, the training optimizer used is excellent for Adam
Change device, initial learning rate is set as 0.0001.Every batch of training data includes 64 40 × 40 image blocks, and training data passes through
Propagated forward is calculated loses with the L2 of clear image, then updates model parameter by the loss backpropagation.Training 100 batches
After secondary, model parameter is saved.
Part is denoised to noise image using the image noise reduction model:
After image noise reduction model training is good, in application, as shown in figure 5, noise image is input to trained image
In noise reduction model, the model parameter kept is loaded, is calculated to front transfer, the noise-free picture after output recovery.
The structure of above-mentioned image noise reduction model is very succinct, and after training, the byte that parameter occupies is seldom, only 0.5M,
Suitable for some non-embedded type mini systems, furthermore, through comparing, above-mentioned gray level image noise-reduction method can rapidly remove image
Noise promotes the visual effect of image.
Technical solution of the present invention and beneficial effect is described in detail in above-described specific embodiment, Ying Li
Solution is not intended to restrict the invention the foregoing is merely presently most preferred embodiment of the invention, all in principle model of the invention
Interior done any modification, supplementary, and equivalent replacement etc. are enclosed, should all be included in the protection scope of the present invention.
Claims (8)
1. a kind of gray level image noise-reduction method based on empty convolution and automatic encoding and decoding neural network, comprising the following steps:
Using clear image and noise image corresponding with clear image as a training sample, training set is constructed with this；
Image noise reduction model is constructed, which includes compiling for carrying out the characteristics of image of feature coding to noise image
The characteristics of image decoding unit that code unit, the coding characteristic figure for exporting to characteristics of image coding unit are decoded, wherein
Characteristics of image coding unit includes characteristic extracting module and 10 coding modules, image decoding unit include 10 decoder modules and
Image restoration module；
Using the image noise reduction model of training set training building, trained image noise reduction model is obtained；
In application, noise image is input to trained image noise reduction model, it is computed output noise-reduced image.
2. the gray level image noise-reduction method based on empty convolution and automatic encoding and decoding neural network as described in claim 1,
It is characterized in that:
Gaussian noise is added to clear image using fixed noise grade, obtains noise image corresponding with clear image.
3. the gray level image noise-reduction method based on empty convolution and automatic encoding and decoding neural network as described in claim 1,
It is characterized in that:
Clear image and noise image corresponding with clear image are divided by several groups image block using identical arbitrary width, with
Every group of image block is as a training sample.
4. the gray level image noise-reduction method based on empty convolution and automatic encoding and decoding neural network as claimed in claim 3,
It is characterized in that:
Clear image and noise image corresponding with clear image are divided into 40 × 40 image block.
5. the gray level image noise-reduction method based on empty convolution and automatic encoding and decoding neural network as described in claim 1,
It is characterized in that:
Characteristic extracting module includes that convolution kernel size is the convolutional layer and active coating that 3 × 3, convolution kernel number is 64；
Coding module include sequentially connected dimensionality reduction submodule, feature extraction submodule and liter dimension submodule, wherein dimensionality reduction
It is the convolutional layer that 1 × 1, convolution kernel number is 32 that submodule and liter dimension submodule, which include a convolution kernel, and connection convolutional layer is defeated
Normalization layer out, the active coating of connection normalization layer output；Feature extraction submodule includes that a convolution kernel is 1 × 1, convolution
Core number is 32, uses expansion rate for the convolutional layer of 2 empty convolution, and the normalization layer of connection convolutional layer output, connection normalizes
The active coating of layer output.
6. the gray level image noise-reduction method based on empty convolution and automatic encoding and decoding neural network as described in claim 1,
It is characterized in that:
Decoder module include sequentially connected dimensionality reduction submodule, feature extraction submodule and liter dimension submodule, wherein dimensionality reduction
It is the warp lamination that 1 × 1, convolution kernel number is 32 that submodule and liter dimension submodule, which include a convolution kernel, connects deconvolution
The normalization layer of layer output, the active coating of connection normalization layer output；Feature extraction submodule include convolution kernel be 1 × 1,
Convolution kernel number is 32, uses expansion rate for the warp lamination of 2 empty convolution, and the normalization layer of connection warp lamination output connects
Connect the active coating of normalization layer output；
Image restoration module includes that a convolution kernel size is the convolutional layer that 3 × 3, convolution kernel number is 64.
7. the gray level image noise-reduction method based on empty convolution and automatic encoding and decoding neural network as described in claim 1,
It is characterized in that:
The activation primitive of active coating is all made of ReLU function.
8. the gray level image noise-reduction method based on empty convolution and automatic encoding and decoding neural network as described in claim 1,
It is characterized in that:
In training process, distorted image successively by characteristics of image coding unit and characteristics of image decoding unit processing after, according to
Clear image calculates the loss of each costing bio disturbance unit, and is weighted superposition to each loss and is finally lost, then
Network weight parameter is updated using final loss backpropagation.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811434831.4A CN109671026B (en) | 2018-11-28 | 2018-11-28 | Gray level image noise reduction method based on void convolution and automatic coding and decoding neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811434831.4A CN109671026B (en) | 2018-11-28 | 2018-11-28 | Gray level image noise reduction method based on void convolution and automatic coding and decoding neural network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109671026A true CN109671026A (en) | 2019-04-23 |
| CN109671026B CN109671026B (en) | 2020-09-29 |
Family
ID=66143280
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811434831.4A Active CN109671026B (en) | 2018-11-28 | 2018-11-28 | Gray level image noise reduction method based on void convolution and automatic coding and decoding neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109671026B (en) |
Cited By (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110163827A (en) * | 2019-05-28 | 2019-08-23 | 腾讯科技（深圳）有限公司 | Training method, image de-noising method, device and the medium of image denoising model |
| CN110290289A (en) * | 2019-06-13 | 2019-09-27 | Oppo广东移动通信有限公司 | Image denoising method, device, electronic equipment and storage medium |
| CN110689496A (en) * | 2019-09-25 | 2020-01-14 | 北京迈格威科技有限公司 | Method and device for determining noise reduction model, electronic equipment and computer storage medium |
| CN110706181A (en) * | 2019-10-09 | 2020-01-17 | 中国科学技术大学 | Image denoising method and system based on multi-scale expansion convolution residual error network |
| CN111429378A (en) * | 2020-04-01 | 2020-07-17 | 天津大学前沿技术研究院有限公司 | Turbid underwater gray level image restoration method based on deep neural network and polarization imaging |
| CN111652186A (en) * | 2020-06-23 | 2020-09-11 | 勇鸿(重庆)信息科技有限公司 | Video category identification method and related device |
| CN112541878A (en) * | 2020-12-24 | 2021-03-23 | 北京百度网讯科技有限公司 | Method and device for establishing image enhancement model and image enhancement |
| CN112784952A (en) * | 2019-11-04 | 2021-05-11 | 珠海格力电器股份有限公司 | Convolutional neural network operation system, method and equipment |
| CN112995673A (en) * | 2019-12-13 | 2021-06-18 | 北京金山云网络技术有限公司 | Sample image processing method and device, electronic equipment and medium |
| CN114118145A (en) * | 2021-11-15 | 2022-03-01 | 北京林业大学 | Method and device for reducing noise of modulation signal, storage medium and equipment |
| US11540798B2 (en) | 2019-08-30 | 2023-01-03 | The Research Foundation For The State University Of New York | Dilated convolutional neural network system and method for positron emission tomography (PET) image denoising |
| CN116016064A (en) * | 2023-01-12 | 2023-04-25 | 西安电子科技大学 | Communication signal noise reduction method based on U-shaped convolution denoising self-encoder |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20160117587A1 (en) * | 2014-10-27 | 2016-04-28 | Zhicheng Yan | Hierarchical deep convolutional neural network for image classification |
| CN106204467A (en) * | 2016-06-27 | 2016-12-07 | 深圳市未来媒体技术研究院 | A kind of image de-noising method based on cascade residual error neutral net |
| 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 |
| CN107229918A (en) * | 2017-05-26 | 2017-10-03 | 西安电子科技大学 | A kind of SAR image object detection method based on full convolutional neural networks |
| CN107240076A (en) * | 2017-05-31 | 2017-10-10 | 北京小米移动软件有限公司 | Image processing method and device |
| WO2018099473A1 (en) * | 2016-12-02 | 2018-06-07 | 北京市商汤科技开发有限公司 | Scene analysis method and system, and electronic device |
| CN108765338A (en) * | 2018-05-28 | 2018-11-06 | 西华大学 | Spatial target images restored method based on convolution own coding convolutional neural networks |
-
2018
- 2018-11-28 CN CN201811434831.4A patent/CN109671026B/en active Active
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20160117587A1 (en) * | 2014-10-27 | 2016-04-28 | Zhicheng Yan | Hierarchical deep convolutional neural network for image classification |
| CN106204467A (en) * | 2016-06-27 | 2016-12-07 | 深圳市未来媒体技术研究院 | A kind of image de-noising method based on cascade residual error neutral net |
| 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 |
| WO2018099473A1 (en) * | 2016-12-02 | 2018-06-07 | 北京市商汤科技开发有限公司 | Scene analysis method and system, and electronic device |
| CN107229918A (en) * | 2017-05-26 | 2017-10-03 | 西安电子科技大学 | A kind of SAR image object detection method based on full convolutional neural networks |
| CN107240076A (en) * | 2017-05-31 | 2017-10-10 | 北京小米移动软件有限公司 | Image processing method and device |
| CN108765338A (en) * | 2018-05-28 | 2018-11-06 | 西华大学 | Spatial target images restored method based on convolution own coding convolutional neural networks |
Cited By (18)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110163827B (en) * | 2019-05-28 | 2023-01-10 | 腾讯科技（深圳）有限公司 | Training method of image denoising model, image denoising method, device and medium |
| CN110163827A (en) * | 2019-05-28 | 2019-08-23 | 腾讯科技（深圳）有限公司 | Training method, image de-noising method, device and the medium of image denoising model |
| CN110290289A (en) * | 2019-06-13 | 2019-09-27 | Oppo广东移动通信有限公司 | Image denoising method, device, electronic equipment and storage medium |
| CN110290289B (en) * | 2019-06-13 | 2021-07-09 | Oppo广东移动通信有限公司 | Image noise reduction method and device, electronic equipment and storage medium |
| US11540798B2 (en) | 2019-08-30 | 2023-01-03 | The Research Foundation For The State University Of New York | Dilated convolutional neural network system and method for positron emission tomography (PET) image denoising |
| CN110689496B (en) * | 2019-09-25 | 2022-10-14 | 北京迈格威科技有限公司 | Method and device for determining noise reduction model, electronic equipment and computer storage medium |
| CN110689496A (en) * | 2019-09-25 | 2020-01-14 | 北京迈格威科技有限公司 | Method and device for determining noise reduction model, electronic equipment and computer storage medium |
| CN110706181A (en) * | 2019-10-09 | 2020-01-17 | 中国科学技术大学 | Image denoising method and system based on multi-scale expansion convolution residual error network |
| CN110706181B (en) * | 2019-10-09 | 2023-04-07 | 中国科学技术大学 | Image denoising method and system based on multi-scale expansion convolution residual error network |
| CN112784952B (en) * | 2019-11-04 | 2024-03-19 | 珠海格力电器股份有限公司 | Convolutional neural network operation system, method and equipment |
| CN112784952A (en) * | 2019-11-04 | 2021-05-11 | 珠海格力电器股份有限公司 | Convolutional neural network operation system, method and equipment |
| CN112995673A (en) * | 2019-12-13 | 2021-06-18 | 北京金山云网络技术有限公司 | Sample image processing method and device, electronic equipment and medium |
| CN112995673B (en) * | 2019-12-13 | 2023-04-07 | 北京金山云网络技术有限公司 | Sample image processing method and device, electronic equipment and medium |
| CN111429378A (en) * | 2020-04-01 | 2020-07-17 | 天津大学前沿技术研究院有限公司 | Turbid underwater gray level image restoration method based on deep neural network and polarization imaging |
| CN111652186A (en) * | 2020-06-23 | 2020-09-11 | 勇鸿(重庆)信息科技有限公司 | Video category identification method and related device |
| CN112541878A (en) * | 2020-12-24 | 2021-03-23 | 北京百度网讯科技有限公司 | Method and device for establishing image enhancement model and image enhancement |
| CN114118145A (en) * | 2021-11-15 | 2022-03-01 | 北京林业大学 | Method and device for reducing noise of modulation signal, storage medium and equipment |
| CN116016064A (en) * | 2023-01-12 | 2023-04-25 | 西安电子科技大学 | Communication signal noise reduction method based on U-shaped convolution denoising self-encoder |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109671026B (en) | 2020-09-29 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109671026A (en) | Gray level image noise-reduction method based on empty convolution and automatic encoding and decoding neural network | |
| CN111832570A (en) | Image semantic segmentation model training method and system | |
| CN110189260A (en) | A kind of image denoising method based on multiple dimensioned parallel gate neural network | |
| CN110443761B (en) | Single image rain removing method based on multi-scale aggregation characteristics | |
| US20230177652A1 (en) | Image restoration method and apparatus, and electronic device | |
| Li et al. | Multi-channel and multi-model-based autoencoding prior for grayscale image restoration | |
| CN113033566A (en) | Model training method, recognition method, device, storage medium, and program product | |
| CN110782406B (en) | Image denoising method and device based on information distillation network | |
| CN113837959B (en) | Image denoising model training method, image denoising method and system | |
| Li et al. | Densely connected network for impulse noise removal | |
| Wang et al. | Learning a multi-level guided residual network for single image deraining | |
| CN110650339A (en) | Video compression method and device and terminal equipment | |
| Li et al. | Deep scale-aware image smoothing | |
| Zeng et al. | Progressive feature fusion attention dense network for speckle noise removal in OCT images | |
| CN112381147A (en) | Dynamic picture similarity model establishing method and device and similarity calculating method and device | |
| CN116703750A (en) | Image defogging method and system based on edge attention and multi-order differential loss | |
| Chen et al. | Image decomposition‐based blind image deconvolution model by employing sparse representation | |
| CN104809749B (en) | A kind of method of the non-sparse blind reconstruction of image compressed sensing | |
| CN112381732B (en) | Image recovery method and system based on multi-scale random proximity algorithm | |
| CN112102216B (en) | Self-adaptive weight total variation image fusion method | |
| CN114529463A (en) | Image denoising method and system | |
| Li et al. | Progressive encoding-decoding image dehazing network | |
| Su et al. | Atmospheric turbulence degraded image restoration using a modified dilated convolutional network | |
| Zhang et al. | Scale-progressive multi-patch network for image dehazing | |
| Zhang et al. | DAResNet Based on double-layer residual block for restoring industrial blurred 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 |