CN108550118A - Fuzzy processing method, device, equipment and the storage medium of motion blur image - Google Patents
Fuzzy processing method, device, equipment and the storage medium of motion blur image Download PDFInfo
- Publication number
- CN108550118A CN108550118A CN201810240092.9A CN201810240092A CN108550118A CN 108550118 A CN108550118 A CN 108550118A CN 201810240092 A CN201810240092 A CN 201810240092A CN 108550118 A CN108550118 A CN 108550118A
- Authority
- CN
- China
- Prior art keywords
- image
- motion blur
- network
- blur image
- fuzzy processing
- 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
- 238000003672 processing method Methods 0.000 title claims abstract description 14
- 238000003860 storage Methods 0.000 title claims abstract description 11
- 238000012545 processing Methods 0.000 claims abstract description 66
- 230000006835 compression Effects 0.000 claims abstract description 35
- 238000007906 compression Methods 0.000 claims abstract description 35
- 230000005284 excitation Effects 0.000 claims abstract description 26
- 238000000605 extraction Methods 0.000 claims abstract description 18
- 238000000034 method Methods 0.000 claims abstract description 15
- 238000012549 training Methods 0.000 claims description 30
- 238000004590 computer program Methods 0.000 claims description 15
- 230000002708 enhancing effect Effects 0.000 claims description 11
- 230000015654 memory Effects 0.000 claims description 7
- 238000002203 pretreatment Methods 0.000 claims description 2
- 230000000694 effects Effects 0.000 abstract description 14
- 238000010586 diagram Methods 0.000 description 5
- 238000004422 calculation algorithm Methods 0.000 description 4
- 230000006870 function Effects 0.000 description 4
- 238000013135 deep learning Methods 0.000 description 3
- 230000008569 process Effects 0.000 description 3
- 238000011084 recovery Methods 0.000 description 3
- 238000011160 research Methods 0.000 description 3
- HUTDUHSNJYTCAR-UHFFFAOYSA-N ancymidol Chemical compound C1=CC(OC)=CC=C1C(O)(C=1C=NC=NC=1)C1CC1 HUTDUHSNJYTCAR-UHFFFAOYSA-N 0.000 description 2
- 239000000203 mixture Substances 0.000 description 2
- 230000008901 benefit Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 230000008602 contraction Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 239000000686 essence Substances 0.000 description 1
- 238000003384 imaging method Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000014759 maintenance of location Effects 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 238000003058 natural language processing Methods 0.000 description 1
- 238000011176 pooling Methods 0.000 description 1
Classifications
-
- G06T5/73—
-
- 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
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/047—Probabilistic or stochastic 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
-
- G06T5/60—
-
- 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]
-
- 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
- G06T2207/20172—Image enhancement details
- G06T2207/20201—Motion blur correction
Abstract
The present invention is applicable in technical field of image processing, provides a kind of fuzzy processing method of motion blur image, device, equipment and storage medium, this method and includes：When receiving the Fuzzy Processing request to motion blur image, the motion blur image is input in the enhanced generator for generating confrontation network trained in advance, the generator includes compression excitation residual error network element and scaling convolution unit, residual error network element is encouraged to carry out feature extraction to motion blur image by compression, to obtain the corresponding characteristic image of motion blur image, Fuzzy Processing is carried out to characteristic image by scaling convolution unit, to obtain the corresponding clear image of motion blur image, the chessboard effect in Fuzzy Processing to reduce motion blur image, improve the clarity of motion blur image restoration, and improve the enhanced Generalization Capability for generating confrontation network of the present invention.
Description
Technical field
The invention belongs to technical field of image processing more particularly to a kind of fuzzy processing method of motion blur image, dresses
It sets, equipment and storage medium.
Background technology
The phenomenon that motion blur is a kind of generally existing in imaging process, in the process of walking, the vapour of aircraft or traveling
When vehicle photographs images, shake can all lead to the generation of motion blur when reference object relative velocity is too fast or shoots, and scheme
The motion blur of picture produces serious influence to the application in fields such as astronomy, military affairs, road traffic, medical images, therefore,
The motion blur removal of image is always one major issue of computer vision field.
As using deep learning algorithm as the emergence of the intelligent algorithm of representative, image procossing, image recognition, language are believed
The research fields such as number processing, natural language processing have obtained development at full speed, and blurred picture is carried out using deep learning algorithm
The research of recovery also achieves some achievements, for example, the generation that the scholars such as Ian Goodfellow in 2014 propose fights network
(GAN), it is become for most very powerful and exceedingly arrogant one of the research direction in deep learning field once proposition.Then, OrestKupyn etc.
Scholar proposed a kind of based on condition confrontation type generation network (CGAN) and content loss (content in 2017 in paper
Loss end-to-end learning method DeblurGAN), the DeblurGAN network models produce on removal image because of object of which movement
Raw blooming is with obvious effects, however, depth of the DeblurGAN network models in removal motion blur image especially image
It will produce " checker-wise shape artifact " (i.e. chessboard effect) after color part, affect the recovery effect of motion blur image.
Invention content
The purpose of the present invention is to provide a kind of fuzzy processing method of motion blur image, device, equipment and storages to be situated between
Matter, it is intended to solve that due to the prior art a kind of fuzzy processing method of effective motion blur image can not be provided, cause right
Motion blur image carries out the problem that chessboard effect is apparent, user experience is bad after Fuzzy Processing.
On the one hand, the present invention provides a kind of fuzzy processing method of motion blur image, the method includes following steps
Suddenly：
When receiving the Fuzzy Processing request to motion blur image, the motion blur image is input to advance instruction
In the generator for the enhanced generation confrontation network perfected, the generator includes that compression excitation residual error network element and scaling are rolled up
Product unit；
Residual error network element is encouraged to carry out feature extraction to the motion blur image by the compression, it is described to obtain
The corresponding characteristic image of motion blur image；
Fuzzy Processing is carried out to the characteristic image by the scaling convolution unit, to obtain the motion blur image
Corresponding clear image.
On the other hand, the present invention provides a kind of Fuzzy Processing device of motion blur image, described device includes：
Image input units, for when receiving to the request of the Fuzzy Processing of motion blur image, by the movement mould
Paste image is input in the enhanced generator for generating confrontation network trained in advance, and the generator includes that compression excitation is residual
Poor network element and scaling convolution unit；
Feature extraction unit, it is special for encouraging residual error network element to carry out the motion blur image by the compression
Sign extraction, to obtain the corresponding characteristic image of the motion blur image；And
Fuzzy Processing unit, for carrying out Fuzzy Processing to the characteristic image by the scaling convolution unit, with
To the corresponding clear image of the motion blur image.
On the other hand, the present invention also provides a kind of computing device, including memory, processor and it is stored in described deposit
In reservoir and the computer program that can run on the processor, the processor are realized such as when executing the computer program
The step of preceding the method.
On the other hand, the present invention also provides a kind of computer readable storage medium, the computer readable storage mediums
It is stored with computer program, the step of computer program realizes method as previously described when being executed by processor.
The motion blur image is input to by the present invention when receiving the Fuzzy Processing request to motion blur image
In the enhanced generator for generating confrontation network trained in advance, which includes compression excitation residual error network element and contracting
Product unit is unreeled, encourages residual error network element to carry out feature extraction to motion blur image by compression, to obtain motion blur
The corresponding characteristic image of image carries out Fuzzy Processing, to obtain motion blur image by scaling convolution unit to characteristic image
Corresponding clear image, the chessboard effect in Fuzzy Processing to reduce motion blur image, improves motion blur figure
As the clarity restored, and improve the enhanced Generalization Capability for generating confrontation network of the present invention.
Description of the drawings
Fig. 1 is the implementation flow chart of the fuzzy processing method for the motion blur image that the embodiment of the present invention one provides；
Fig. 2 be the embodiment of the present invention one provide motion blur image fuzzy processing method in compression excitation residual error network
Unit carries out the flow diagram of feature extraction；
Fig. 3 is the structural schematic diagram of the Fuzzy Processing device of motion blur image provided by Embodiment 2 of the present invention；
Fig. 4 is the structural schematic diagram of the Fuzzy Processing device for the motion blur image that the embodiment of the present invention three provides；And
Fig. 5 is the structural schematic diagram for the computing device that the embodiment of the present invention four provides.
Specific implementation mode
In order to make the purpose , technical scheme and advantage of the present invention be clearer, with reference to the accompanying drawings and embodiments, right
The present invention is further elaborated.It should be appreciated that the specific embodiments described herein are merely illustrative of the present invention, and
It is not used in the restriction present invention.
The specific implementation of the present invention is described in detail below in conjunction with specific embodiment：
Embodiment one：
Fig. 1 shows the implementation process of the fuzzy processing method for the motion blur image that the embodiment of the present invention one provides, and is
Convenient for explanation, illustrate only with relevant part of the embodiment of the present invention, details are as follows：
In step S101, when receiving the Fuzzy Processing request to motion blur image, by the motion blur image
It is input in the enhanced generator for generating confrontation network trained in advance, generator includes compression excitation residual error network element
With scaling convolution unit.
The embodiment of the present invention is suitable for computing device, for example, personal computer, smart mobile phone, tablet etc..By the movement
Blurred picture is input to before the enhanced generator for generating confrontation network (EDGAN) trained in advance, it is preferable that by pre-
If formatting pretreatment layer pretreatment is formatted to motion blur image, to by formatting pretreated motion blur
Image carries out image enhancement processing, and the motion blur image enhanced fights to improve subsequently through enhanced generation
Network carries out the motion blur image effect of Fuzzy Processing.
It is further preferred that preset formatting pretreatment layer is using three-dimensional Block- matching (Block-Matching3D, abbreviation
BM3D) either likelihood probability logarithm it is expected (Expected Patch Log Likelihood abbreviation EPLL) or Weighted Kernel model
Number minimizes (WeightedNuclearNorm Minimization, abbreviation WNNM) Denoising Algorithm and is carried out to motion blur image
Pretreatment is formatted, to carry out denoising to the motion blur image, to improve minutia in the motion blur image
Obvious degree.
Another is preferably to format pretreatment layer by the enhanced generation confrontation network settings trained in advance, right
Motion blur image is formatted pretreatment, to carry out denoising to the motion blur image, to improve to motion blur
The effect that picture noise filters out.
It is further preferred that when to carrying out image enhancement processing by formatting pretreated motion blur image, it can
To be zoomed in and out to image by preset resolution contraction and enlargement rate, random cropping then is carried out to the image after scaling, with
To the motion blur image of enhancing, for example, the blurred picture of 2560 × 720 resolution ratio is zoomed into 640 × 360 resolution ratio, then
Random cropping is 256 × 256 image in different resolution, to improve the enhanced training effect for generating confrontation network of the embodiment of the present invention
Rate, and improve the efficiency that Fuzzy Processing is subsequently carried out to motion blur image.
When receiving the Fuzzy Processing request to motion blur image, which is input to advance training
Before in the generator of good enhanced generation confrontation network, it is another preferably, structure two generates confrontation networks in advance, this two
A generation confrontation network fights network for enhanced generation, and each confrontation network that generates swashs comprising two convolution units, 9 compressions
Residual error network element and two scaling convolution units are encouraged, to improve the Generalization Capability of network, by generating confrontation net to two
The application enhanced generation confrontation network trained in advance can be obtained in the training of network.For ease of description, herein by this two
A generation confrontation network is denoted as the first generation confrontation network and second and generates confrontation network.
It is further preferred that when building two generators for generating confrontation network respectively, remove all in generator
Norm layers of Batch Norm or Instance to improve the training speed of network and the stability of network, and then are avoided brokenly
The contrast information of bad image script.
After initial construction above-mentioned two generates confrontation network, generates confrontation network to two and be trained, to obtain
Enhanced generation trained in advance fights network.Preferably, it is trained, to two generation confrontation networks to obtain in advance
When trained enhanced generation confrontation network, the first training sample is input to first built in advance and is generated in confrontation network
It is trained, confrontation network is generated to obtain trained first, generating confrontation network pair second by trained first instructs
Practice sample and is formatted pretreatment, to carrying out image enhancement processing by formatting pretreated second training sample, with
To the second training sample of enhancing, the second training sample of enhancing is input to build in advance second generate confrontation network in into
Row training generates confrontation network to obtain trained second, is enhancing by the trained second generation confrontation network settings
Type generates confrontation network (i.e.：Enhanced generation trained in advance fights network).By the trained increasing of the embodiment of the present invention
Strong type generates confrontation network and carries out Fuzzy Processing to motion blur image, improves the peak value noise of the motion blur image of recovery
Than (PSNR) and structural similarity (SSIM).
In step s 102, residual error network element is encouraged to carry out feature extraction to motion blur image by compression, with
To the corresponding characteristic image of motion blur image.
In embodiments of the present invention, it is preferable that compression excitation residual error network element is by residual error network and compression excitation network
Composition, the residual error network do not include Norm layers of Batch Norm or Instance, which includes that an overall situation is flat
Equal pond layer, 2 full articulamentums, one Relu layers and one Sigmoid layers, to improve the performance of feature extraction.
As illustratively, Fig. 2 is the flow diagram that compression excitation residual error network element carries out feature extraction, is input to pressure
The image of contracting excitation residual error network element (SE-ResBlock) is first 3 by two convolution kernel sizes in residual error network
After × 3 convolutional layer (3 × 3Conv) convolution sum is connected to the excitation layer Relu operations between two convolutional layers, obtain preliminary
Characteristic image, then by this feature image be input to compression excitation network the overall situation be averaged pond layer (Global Pooling) progress
Characteristic image through overcompression is input to a full articulamentum (FC) by squeeze operation, and the full articulamentum is by the spy of characteristic image
Sign dimension is reduced to the 1/16 of input, then, again by a full articulamentum (FC) by characteristic pattern after Relu is activated
The characteristic dimension of picture rises back original dimension, passes through a Sigmoid later and encourages, and will return finally by a Scale operation
On Weight to each channel characteristics after one change, the corresponding characteristic image of final output motion blur image, to improve
The performance of feature extraction.
In step s 103, Fuzzy Processing is carried out to characteristic image by scaling convolution unit, to obtain motion blur figure
As corresponding clear image.
In embodiments of the present invention, when carrying out Fuzzy Processing to characteristic image by scaling convolution unit, it is preferable that first
First, it is contracted according to preset ratio to the size of this feature image by preset arest neighbors interpolation or bilinear interpolation
It puts, then, the characteristic image Jing Guo scaling processing is up-sampled by convolution operation, it is corresponding clear to obtain this feature image
Clear image, the chessboard effect in the clear image to reduce, so as to get clear image can be in higher resolution
Display equipment on clearly show.
As illustratively, table 1 is shown through GOPRO data sets and Lai data sets in DebluGAN networks and the present invention
The Y-PSNR (PSNR) tested and structural similarity (SSIM) under the enhanced generation confrontation network of embodiment, can be with
Find out that the enhanced generation obtained Y-PSNR of confrontation network and structural similarity ratio through the embodiment of the present invention pass through
What DebluGAN networks obtained is significantly improved.
Table 1
In embodiments of the present invention, when receiving the Fuzzy Processing request to motion blur image, by the motion blur
Image is input in the enhanced generator for generating confrontation network trained in advance, which includes compression excitation residual error net
Network unit and scaling convolution unit, encourage residual error network element to carry out feature extraction to motion blur image by compression, with
To the corresponding characteristic image of motion blur image, Fuzzy Processing is carried out to characteristic image by scaling convolution unit, to be transported
The corresponding clear image of dynamic blurred picture, the chessboard effect in Fuzzy Processing to reduce motion blur image improve
The clarity of motion blur image restoration, and improve the enhanced Generalization Capability for generating confrontation network of the embodiment of the present invention.
Embodiment two：
Fig. 3 shows the structure of the Fuzzy Processing device of motion blur image provided by Embodiment 2 of the present invention, in order to just
In explanation, illustrate only with the relevant part of the embodiment of the present invention, including：
Image input units 31, for when receiving to the request of the Fuzzy Processing of motion blur image, by the movement mould
Paste image is input in the enhanced generator for generating confrontation network trained in advance, and generator includes compression excitation residual error net
Network unit and scaling convolution unit.
The embodiment of the present invention is suitable for computing device, for example, personal computer, smart mobile phone, tablet etc..When receiving pair
When the Fuzzy Processing request of motion blur image, which is input to enhanced generation trained in advance and is fought
Before in the generator of network, it is preferable that build enhanced generation confrontation network in advance, which includes
Two convolution units, 9 compression excitation residual error network element and two scaling convolution units, to improve generalization ability of network
Energy.
It is further preferred that when building the enhanced generator for generating confrontation network, remove all in generator
Norm layers of Batch Norm or Instance to improve the training speed of network and the stability of network, and then are avoided
Destroy the contrast information of image script.
Feature extraction unit 32 puies forward motion blur image progress feature for passing through compression excitation residual error network element
It takes, to obtain the corresponding characteristic image of motion blur image.
In embodiments of the present invention, it is preferable that compression excitation residual error network element is by residual error network and compression excitation network
Composition, the residual error network do not include Norm layers of Batch Norm or Instance, which includes that an overall situation is flat
Equal pond layer, 2 full articulamentums, one Relu layers and one Sigmoid layers, to improve the performance of feature extraction.
Fuzzy Processing unit 33, for carrying out Fuzzy Processing to characteristic image by scaling convolution unit, to be moved
The corresponding clear image of blurred picture.
In embodiments of the present invention, when carrying out Fuzzy Processing to characteristic image by scaling convolution unit, it is preferable that first
First, it is contracted according to preset ratio to the size of this feature image by preset arest neighbors interpolation or bilinear interpolation
It puts, then, the characteristic image Jing Guo scaling processing is up-sampled by convolution operation, it is corresponding clear to obtain this feature image
Clear image, the chessboard effect in the clear image to reduce, so as to get clear image can be in higher resolution
Display equipment on clearly show.
In embodiments of the present invention, each unit of the Fuzzy Processing device of motion blur image can be by corresponding hardware or soft
Part unit realizes that each unit can be independent soft and hardware unit, can also be integrated into a soft and hardware unit, not have to herein
To limit the present invention.
Embodiment three：
Fig. 4 shows the structure of the Fuzzy Processing device for the motion blur image that the embodiment of the present invention three provides, in order to just
In explanation, illustrate only with the relevant part of the embodiment of the present invention, including：
First network training unit 41 generates confrontation network for the first training sample to be input to build in advance first
In be trained, with obtain it is trained first generate confrontation network；
Sample pre-treatment unit 42 is formatted pre- place for passing through the first generation confrontation the second training sample of network pair
Reason；
Sample enhancement unit 43, for formatting pretreated second training sample progress image enhancement processing to passing through,
With the second training sample enhanced；
Second network training unit 44, the second training sample for that will enhance are input to the second generation pair built in advance
It is trained in anti-network, confrontation network is generated to obtain trained second, by the trained second generation confrontation network
It is set as enhanced generation confrontation network (i.e.：Enhanced generation trained in advance fights network)；
Pretreatment unit 45 is formatted, for when receiving to the request of the Fuzzy Processing of motion blur image, by pre-
If formatting pretreatment layer pretreatment is formatted to motion blur image；
Image enhancing unit 46, for formatting pretreated motion blur image progress image enhancement processing to passing through,
The motion blur image enhanced；
Image input units 47, for the motion blur image of enhancing to be input to enhanced generation pair trained in advance
In the generator of anti-network, generator includes compression excitation residual error network element and scaling convolution unit；
Feature extraction unit 48 puies forward motion blur image progress feature for passing through compression excitation residual error network element
It takes, to obtain the corresponding characteristic image of motion blur image；And
Fuzzy Processing unit 49, for carrying out Fuzzy Processing to characteristic image by scaling convolution unit, to be moved
The corresponding clear image of blurred picture.
In embodiments of the present invention, each unit of the Fuzzy Processing device of motion blur image can be by corresponding hardware or soft
Part unit realizes that each unit can be independent soft and hardware unit, can also be integrated into a soft and hardware unit, not have to herein
To limit the present invention.The specific implementation mode of each unit can refer to the description of embodiment one, and details are not described herein.
Example IV：
Fig. 5 shows the structure for the computing device that the embodiment of the present invention four provides, and for convenience of description, illustrates only and this
The relevant part of inventive embodiments.
The computing device 5 of the embodiment of the present invention includes processor 50, memory 51 and is stored in memory 51 and can
The computer program 52 run on processor 50.The processor 50 realizes above-mentioned motion blur figure when executing computer program 52
Step in the fuzzy processing method embodiment of picture, such as step S101 to S103 shown in FIG. 1.Alternatively, processor 50 executes
The function of each unit in above-mentioned each device embodiment, such as the function of unit 31 to 33 shown in Fig. 3 are realized when computer program 52.
In embodiments of the present invention, when receiving the Fuzzy Processing request to motion blur image, by the motion blur
Image is input in the enhanced generator for generating confrontation network trained in advance, which includes compression excitation residual error net
Network unit and scaling convolution unit, encourage residual error network element to carry out feature extraction to motion blur image by compression, with
To the corresponding characteristic image of motion blur image, Fuzzy Processing is carried out to characteristic image by scaling convolution unit, to be transported
The corresponding clear image of dynamic blurred picture, the chessboard effect in Fuzzy Processing to reduce motion blur image improve
The clarity of motion blur image restoration, and improve the enhanced Generalization Capability for generating confrontation network of the embodiment of the present invention.
The computing device of the embodiment of the present invention can be personal computer, smart mobile phone, tablet.In the computing device 5
The step of reason device 50 is realized when realizing the fuzzy processing method of motion blur image when executing computer program 52 can refer to aforementioned
The description of embodiment of the method, details are not described herein.
Embodiment five：
In embodiments of the present invention, a kind of computer readable storage medium is provided, which deposits
Computer program is contained, which realizes that the fuzzy processing method of above-mentioned motion blur image is real when being executed by processor
The step in example is applied, for example, step S101 to S103 shown in FIG. 1.Alternatively, the computer program is realized when being executed by processor
The function of each unit in above-mentioned each device embodiment, for example, unit 31 to 33 shown in Fig. 3 function.
In embodiments of the present invention, when receiving the Fuzzy Processing request to motion blur image, by the motion blur
Image is input in the enhanced generator for generating confrontation network trained in advance, which includes compression excitation residual error net
Network unit and scaling convolution unit, encourage residual error network element to carry out feature extraction to motion blur image by compression, with
To the corresponding characteristic image of motion blur image, Fuzzy Processing is carried out to characteristic image by scaling convolution unit, to be transported
The corresponding clear image of dynamic blurred picture, the chessboard effect in Fuzzy Processing to reduce motion blur image improve
The clarity of motion blur image restoration, and improve the enhanced Generalization Capability for generating confrontation network of the embodiment of the present invention.
The computer readable storage medium of the embodiment of the present invention may include can carry computer program code any
Entity or device, recording medium, for example, the memories such as ROM/RAM, disk, CD, flash memory.
The foregoing is merely illustrative of the preferred embodiments of the present invention, is not intended to limit the invention, all essences in the present invention
All any modification, equivalent and improvement etc., should all be included in the protection scope of the present invention made by within refreshing and principle.
Claims (10)
1. a kind of fuzzy processing method of motion blur image, which is characterized in that the method includes following step：
When receiving the Fuzzy Processing request to motion blur image, the motion blur image is input to and is trained in advance
Enhanced generation confrontation network generator in, the generator include compression excitation residual error network element and scaling convolution list
Member；
Residual error network element is encouraged to carry out feature extraction to the motion blur image by the compression, to obtain the movement
The corresponding characteristic image of blurred picture；
Fuzzy Processing is carried out to the characteristic image by the scaling convolution unit, is corresponded to obtaining the motion blur image
Clear image.
2. the method as described in claim 1, which is characterized in that the motion blur image is input to advance trained increasing
Before strong type generates the step in the generator of confrontation network, including：
When receiving the Fuzzy Processing request to motion blur image, by preset formatting pretreatment layer to the movement
Blurred picture is formatted pretreatment；
To described image enhancement processing, the movement mould enhanced are carried out by formatting pretreated motion blur image
Paste image.
3. the method as described in claim 1, which is characterized in that the compression encourages residual error network element by residual error network and pressure
Contracting excitation network forms.
4. the method as described in claim 1, which is characterized in that asked when receiving the Fuzzy Processing to motion blur image
When, by the motion blur image be input in advance it is trained it is enhanced generate confrontation network generator in step it
Before, including：
First training sample is input in the first generation confrontation network built in advance and is trained, to obtain trained institute
State the first generation confrontation network；
Confrontation the second training sample of network pair, which is generated, by described first is formatted pretreatment；
Carry out image enhancement processing by formatting pretreated second training sample to described, with enhanced described second
Training sample；
Second training sample of enhancing is input in the second generation confrontation network built in advance and is trained, to obtain
Trained described second generates confrontation network, and it is the enhancing to generate confrontation network settings by described trained described second
Type generates confrontation network.
5. a kind of Fuzzy Processing device of motion blur image, which is characterized in that described device includes：
Image input units, for when receiving to the request of the Fuzzy Processing of motion blur image, by the motion blur figure
As being input in the enhanced generator for generating confrontation network trained in advance, the generator includes compression excitation residual error net
Network unit and scaling convolution unit；
Feature extraction unit is carried for encouraging residual error network element to carry out feature to the motion blur image by the compression
It takes, to obtain the corresponding characteristic image of the motion blur image；And
Fuzzy Processing unit, for carrying out Fuzzy Processing to the characteristic image by the scaling convolution unit, to obtain
State the corresponding clear image of motion blur image.
6. device as claimed in claim 5, which is characterized in that described device further includes：
Pretreatment unit is formatted, for when receiving the Fuzzy Processing request to motion blur image, passing through preset lattice
Formula pretreatment layer is formatted pretreatment to the motion blur image；And
Image enhancing unit is obtained for carrying out image enhancement processing by formatting pretreated motion blur image to described
To the motion blur image of enhancing.
7. device as claimed in claim 5, which is characterized in that the compression encourages residual error network element by residual error network and pressure
Contracting excitation network forms.
8. device as claimed in claim 5, which is characterized in that described device further includes：
First network training unit is carried out for being input to the first training sample in the build in advance first generation confrontation network
Training generates confrontation network to obtain trained described first；
Sample pre-treatment unit is formatted pre- place for passing through the first generation confrontation second training sample of network pair
Reason；
Sample enhancement unit, for carrying out image enhancement processing by formatting pretreated second training sample to described, with
Second training sample enhanced；And
Second network training unit generates confrontation for second training sample of enhancing to be input to build in advance second
It is trained in network, generates confrontation network to obtain trained described second, described trained described second is generated
It is that the enhanced generation fights network to fight network settings.
9. a kind of computing device, including memory, processor and it is stored in the memory and can be on the processor
The computer program of operation, which is characterized in that the processor realizes such as Claims 1-4 when executing the computer program
The step of any one the method.
10. a kind of computer readable storage medium, the computer-readable recording medium storage has computer program, feature to exist
In when the computer program is executed by processor the step of any one of such as Claims 1-4 of realization the method.
Priority Applications (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810240092.9A CN108550118B (en) | 2018-03-22 | 2018-03-22 | Motion blur image blur processing method, device, equipment and storage medium |
| PCT/CN2018/081710 WO2019178893A1 (en) | 2018-03-22 | 2018-04-03 | Motion blur image sharpening method and device, apparatus, and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810240092.9A CN108550118B (en) | 2018-03-22 | 2018-03-22 | Motion blur image blur processing method, device, equipment and storage medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108550118A true CN108550118A (en) | 2018-09-18 |
| CN108550118B CN108550118B (en) | 2022-02-22 |
Family
ID=63516720
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810240092.9A Expired - Fee Related CN108550118B (en) | 2018-03-22 | 2018-03-22 | Motion blur image blur processing method, device, equipment and storage medium |
Country Status (2)
| Country | Link |
|---|---|
| CN (1) | CN108550118B (en) |
| WO (1) | WO2019178893A1 (en) |
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109712092A (en) * | 2018-12-18 | 2019-05-03 | 上海中信信息发展股份有限公司 | Archives scan image repair method, device and electronic equipment |
| CN109859141A (en) * | 2019-02-18 | 2019-06-07 | 安徽理工大学 | A kind of depth shaft wall image de-noising method |
| CN110012145A (en) * | 2019-04-08 | 2019-07-12 | 北京易诚高科科技发展有限公司 | A kind of mobile phone stabilization function evaluating method based on image blur |
| CN110070517A (en) * | 2019-03-14 | 2019-07-30 | 安徽艾睿思智能科技有限公司 | Blurred picture synthetic method based on degeneration imaging mechanism and generation confrontation mechanism |
| CN110074813A (en) * | 2019-04-26 | 2019-08-02 | 深圳大学 | A kind of ultrasonic image reconstruction method and system |
| CN111340716A (en) * | 2019-11-20 | 2020-06-26 | 电子科技大学成都学院 | Image deblurring method for improving dual-discrimination countermeasure network model |
| CN111768826A (en) * | 2020-06-30 | 2020-10-13 | 平安国际智慧城市科技股份有限公司 | Electronic health case generation method and device, terminal equipment and storage medium |
| CN112598593A (en) * | 2020-12-25 | 2021-04-02 | 吉林大学 | Seismic noise suppression method based on non-equilibrium depth expectation block log-likelihood network |
| CN112651894A (en) * | 2020-12-29 | 2021-04-13 | 湖北工业大学 | Image deblurring algorithm for generating countermeasure network based on RRDB |
| CN112949460A (en) * | 2021-02-26 | 2021-06-11 | 陕西理工大学 | Human body behavior network model based on video and identification method |
| CN113313180A (en) * | 2021-06-04 | 2021-08-27 | 太原理工大学 | Remote sensing image semantic segmentation method based on deep confrontation learning |
Families Citing this family (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112904548B (en) * | 2019-12-03 | 2023-06-09 | 精微视达医疗科技（武汉）有限公司 | Endoscope focusing method and device |
| CN111275637B (en) * | 2020-01-15 | 2024-01-30 | 北京工业大学 | Attention model-based non-uniform motion blurred image self-adaptive restoration method |
| CN111369451B (en) * | 2020-02-24 | 2023-08-01 | 黑蜂智造(深圳)科技有限公司 | Image restoration model, method and device based on complex task decomposition regularization |
| CN111460939A (en) * | 2020-03-20 | 2020-07-28 | 深圳市优必选科技股份有限公司 | Deblurring face recognition method and system and inspection robot |
| CN112435179A (en) * | 2020-11-11 | 2021-03-02 | 北京工业大学 | Fuzzy pollen particle picture processing method and device and electronic equipment |
| CN112419201A (en) * | 2020-12-04 | 2021-02-26 | 珠海亿智电子科技有限公司 | Image deblurring method based on residual error network |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104091340A (en) * | 2014-07-18 | 2014-10-08 | 厦门美图之家科技有限公司 | Blurred image rapid detection method |
| CN105118031A (en) * | 2015-08-11 | 2015-12-02 | 中国科学院计算技术研究所 | Image processing method for recovering depth information |
| 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 |
| CN106952239A (en) * | 2017-03-28 | 2017-07-14 | 厦门幻世网络科技有限公司 | image generating method and device |
| CN107527044A (en) * | 2017-09-18 | 2017-12-29 | 北京邮电大学 | A kind of multiple car plate clarification methods and device based on search |
| CN107590774A (en) * | 2017-09-18 | 2018-01-16 | 北京邮电大学 | A kind of car plate clarification method and device based on generation confrontation network |
| CN107609560A (en) * | 2017-09-27 | 2018-01-19 | 北京小米移动软件有限公司 | Character recognition method and device |
| CN107730458A (en) * | 2017-09-05 | 2018-02-23 | 北京飞搜科技有限公司 | A kind of fuzzy facial reconstruction method and system based on production confrontation network |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107330954A (en) * | 2017-07-14 | 2017-11-07 | 深圳市唯特视科技有限公司 | A kind of method based on attenuation network by sliding attribute manipulation image |
-
2018
- 2018-03-22 CN CN201810240092.9A patent/CN108550118B/en not_active Expired - Fee Related
- 2018-04-03 WO PCT/CN2018/081710 patent/WO2019178893A1/en active Application Filing
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104091340A (en) * | 2014-07-18 | 2014-10-08 | 厦门美图之家科技有限公司 | Blurred image rapid detection method |
| CN105118031A (en) * | 2015-08-11 | 2015-12-02 | 中国科学院计算技术研究所 | Image processing method for recovering depth information |
| 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 |
| CN106952239A (en) * | 2017-03-28 | 2017-07-14 | 厦门幻世网络科技有限公司 | image generating method and device |
| CN107730458A (en) * | 2017-09-05 | 2018-02-23 | 北京飞搜科技有限公司 | A kind of fuzzy facial reconstruction method and system based on production confrontation network |
| CN107527044A (en) * | 2017-09-18 | 2017-12-29 | 北京邮电大学 | A kind of multiple car plate clarification methods and device based on search |
| CN107590774A (en) * | 2017-09-18 | 2018-01-16 | 北京邮电大学 | A kind of car plate clarification method and device based on generation confrontation network |
| CN107609560A (en) * | 2017-09-27 | 2018-01-19 | 北京小米移动软件有限公司 | Character recognition method and device |
Non-Patent Citations (1)
| Title |
|---|
| 刘磊: "深度图像下基于头部多特征的人数统计算法", 《深圳大学学报(理工版)》 * |
Cited By (17)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109712092A (en) * | 2018-12-18 | 2019-05-03 | 上海中信信息发展股份有限公司 | Archives scan image repair method, device and electronic equipment |
| CN109859141A (en) * | 2019-02-18 | 2019-06-07 | 安徽理工大学 | A kind of depth shaft wall image de-noising method |
| CN109859141B (en) * | 2019-02-18 | 2022-05-27 | 安徽理工大学 | Deep vertical shaft well wall image denoising method |
| CN110070517B (en) * | 2019-03-14 | 2021-05-25 | 安徽艾睿思智能科技有限公司 | Blurred image synthesis method based on degradation imaging mechanism and generation countermeasure mechanism |
| CN110070517A (en) * | 2019-03-14 | 2019-07-30 | 安徽艾睿思智能科技有限公司 | Blurred picture synthetic method based on degeneration imaging mechanism and generation confrontation mechanism |
| CN110012145A (en) * | 2019-04-08 | 2019-07-12 | 北京易诚高科科技发展有限公司 | A kind of mobile phone stabilization function evaluating method based on image blur |
| CN110074813B (en) * | 2019-04-26 | 2022-03-04 | 深圳大学 | Ultrasonic image reconstruction method and system |
| CN110074813A (en) * | 2019-04-26 | 2019-08-02 | 深圳大学 | A kind of ultrasonic image reconstruction method and system |
| CN111340716A (en) * | 2019-11-20 | 2020-06-26 | 电子科技大学成都学院 | Image deblurring method for improving dual-discrimination countermeasure network model |
| CN111768826A (en) * | 2020-06-30 | 2020-10-13 | 平安国际智慧城市科技股份有限公司 | Electronic health case generation method and device, terminal equipment and storage medium |
| CN111768826B (en) * | 2020-06-30 | 2023-06-27 | 深圳平安智慧医健科技有限公司 | Electronic health case generation method, device, terminal equipment and storage medium |
| CN112598593A (en) * | 2020-12-25 | 2021-04-02 | 吉林大学 | Seismic noise suppression method based on non-equilibrium depth expectation block log-likelihood network |
| CN112598593B (en) * | 2020-12-25 | 2022-05-27 | 吉林大学 | Seismic noise suppression method based on non-equilibrium depth expectation block log-likelihood network |
| CN112651894A (en) * | 2020-12-29 | 2021-04-13 | 湖北工业大学 | Image deblurring algorithm for generating countermeasure network based on RRDB |
| CN112949460A (en) * | 2021-02-26 | 2021-06-11 | 陕西理工大学 | Human body behavior network model based on video and identification method |
| CN112949460B (en) * | 2021-02-26 | 2024-02-13 | 陕西理工大学 | Human behavior network model based on video and identification method |
| CN113313180A (en) * | 2021-06-04 | 2021-08-27 | 太原理工大学 | Remote sensing image semantic segmentation method based on deep confrontation learning |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108550118B (en) | 2022-02-22 |
| WO2019178893A1 (en) | 2019-09-26 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108550118A (en) | Fuzzy processing method, device, equipment and the storage medium of motion blur image | |
| Hai et al. | R2rnet: Low-light image enhancement via real-low to real-normal network | |
| Nah et al. | Deep multi-scale convolutional neural network for dynamic scene deblurring | |
| Sun et al. | Lightweight image super-resolution via weighted multi-scale residual network | |
| Yu et al. | A unified learning framework for single image super-resolution | |
| CN107392852B (en) | Super-resolution reconstruction method, device and equipment for depth image and storage medium | |
| Zheng et al. | Learning frequency domain priors for image demoireing | |
| CN110060204B (en) | Single image super-resolution method based on reversible network | |
| CN111091503A (en) | Image out-of-focus blur removing method based on deep learning | |
| Min et al. | Blind deblurring via a novel recursive deep CNN improved by wavelet transform | |
| Yang et al. | Ensemble learning priors driven deep unfolding for scalable video snapshot compressive imaging | |
| Fang et al. | High-resolution optical flow and frame-recurrent network for video super-resolution and deblurring | |
| Zheng et al. | T-net: Deep stacked scale-iteration network for image dehazing | |
| Chen et al. | Image denoising via deep network based on edge enhancement | |
| Pang et al. | Lightweight multi-scale aggregated residual attention networks for image super-resolution | |
| CN116523800A (en) | Image noise reduction model and method based on residual dense network and attention mechanism | |
| Zhu et al. | Enlightening low-light images with dynamic guidance for context enrichment | |
| Li et al. | High-resolution network for photorealistic style transfer | |
| Guo et al. | Image blind deblurring using an adaptive patch prior | |
| Jiang et al. | Image motion deblurring based on deep residual shrinkage and generative adversarial networks | |
| CN116485654A (en) | Lightweight single-image super-resolution reconstruction method combining convolutional neural network and transducer | |
| CN116266336A (en) | Video super-resolution reconstruction method, device, computing equipment and storage medium | |
| Pang et al. | Image super-resolution based on generalized residual network | |
| Kim et al. | Light field angular super-resolution using convolutional neural network with residual network | |
| Li et al. | Deep ranking exemplar-based dynamic scene deblurring |
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 | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20220222 |
|
| CF01 | Termination of patent right due to non-payment of annual fee |