CN109636746A - Picture noise removes system, method and apparatus - Google Patents
Picture noise removes system, method and apparatus Download PDFInfo
- Publication number
- CN109636746A CN109636746A CN201811459206.5A CN201811459206A CN109636746A CN 109636746 A CN109636746 A CN 109636746A CN 201811459206 A CN201811459206 A CN 201811459206A CN 109636746 A CN109636746 A CN 109636746A
- Authority
- CN
- China
- Prior art keywords
- noise
- image
- network
- denoising
- picture
- 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 32
- 238000012549 training Methods 0.000 claims abstract description 33
- 230000004069 differentiation Effects 0.000 claims abstract description 6
- 230000006870 function Effects 0.000 claims description 28
- 239000011159 matrix material Substances 0.000 claims description 12
- 230000008030 elimination Effects 0.000 claims description 9
- 238000003379 elimination reaction Methods 0.000 claims description 9
- 238000010801 machine learning Methods 0.000 claims description 9
- 238000004422 calculation algorithm Methods 0.000 claims description 8
- 238000013528 artificial neural network Methods 0.000 claims description 7
- 238000010030 laminating Methods 0.000 claims description 7
- 235000013399 edible fruits Nutrition 0.000 claims 1
- 238000012545 processing Methods 0.000 abstract description 4
- 238000010586 diagram Methods 0.000 description 9
- 230000009467 reduction Effects 0.000 description 4
- 230000008569 process Effects 0.000 description 3
- 230000009286 beneficial effect Effects 0.000 description 2
- 230000008901 benefit Effects 0.000 description 2
- 230000008859 change Effects 0.000 description 2
- 238000005242 forging Methods 0.000 description 2
- 230000001537 neural effect Effects 0.000 description 2
- 238000004088 simulation Methods 0.000 description 2
- 230000009471 action Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 201000010099 disease Diseases 0.000 description 1
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000004927 fusion Effects 0.000 description 1
- 238000012806 monitoring device Methods 0.000 description 1
- 210000005036 nerve Anatomy 0.000 description 1
- 230000000717 retained effect Effects 0.000 description 1
Classifications
-
- G06T5/70—
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10072—Tomographic images
- G06T2207/10081—Computed x-ray tomography [CT]
-
- 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]
Abstract
The invention belongs to technical field of image processing, specifically provide a kind of picture noise removal system, method and apparatus, it is intended to solve the problems, such as that the prior art has noise and muting training data to realize removal noise in image based on what picture noise distributional assumption and needs were matched.For this purpose, picture noise removal system provided by the invention includes generation confrontation network and image denoising network, generating confrontation network includes that noise generates network and noise differentiation network；Image denoising network configuration is to carry out denoising to target image to obtain low noise image；It is to generate noise image according to low noise image sample and random noise that noise, which generates network configuration,；Noise differentiates that network configuration is that the probability that the noise image that network exports is real noise image is generated according to preset noise image sample predictions noise.Based on above structure, system provided by the invention can be adapted for various types of noises, and more economical.
Description
Technical field
The invention belongs to technical field of image processing, and in particular to a kind of picture noise removal system, method and apparatus.
Background technique
When handling image data, what noise data was almost constantly present.Either medical image, mobile phone shooting
Photo, or monitoring device image more or less include noise data, remove picture noise data for subsequent figure
As processing is essential processing step.
The method of existing removal picture noise mainly includes two classes:
The first kind assumes that the distribution situation of noise, by taking medical CT image as an example, usually assumes that making an uproar in medical CT image
Sound obeys Poisson distribution, and the noise of different location is independent from each other, but in practical application, it is assumed that noise profile situation
In real data and unreasonable, the noise of different location is likely to very strong correlation in image, different images
Noise profile situation is different, cannot reflect true situation to the hypothesis of noise；
Second class is trained neural network in advance will to be inputted with noisy image, output is not based on machine learning
Image with noise, but the training data of training neural network needs to be containing the noise and not image of Noise matched
Data, and the training data matched is generally difficult to obtain generally use artificial data in practical study, but artificial number
According to the noise profile situation for not necessarily complying fully with data.By taking medical image as an example, although can be to the same object or disease
People uses different radiant quantities, reaches different noise levels, but is similarly difficult to realize the image complete one shot twice
It causes, and repeatedly shooting cost is also a very big obstacle.
Therefore, how to propose that a kind of be both distributed to picture noise had not been done any it is assumed that and having noise and nothing without pairing
The training data of noise realizes that the scheme of removal noise in image data is the current problem to be solved of those skilled in the art.
Summary of the invention
In order to solve the above problem in the prior art, in order to solve the prior art be based on picture noise distributional assumption with
And need to match has the problem of noise and muting training data realize removal noise in image, the first aspect of the present invention
Provide a kind of picture noise removal system, comprising: generate confrontation network and image denoising network, the generation fights network packet
It includes noise and generates network and noise differentiation network；
Described image denoising network configuration obtains low noise image to carry out denoising to target image, or described
System generates the defeated of network to the noise when carrying out network training to the generation confrontation network and described image denoising network
Result carries out denoising and obtains low noise image sample out；
It is to generate noise image simultaneously according to the low noise image sample and random noise that the noise, which generates network configuration,
And the noise image is exported respectively to described image and denoises network and noise differentiation network；
The noise differentiates that network configuration is that the noise according to preset noise image sample predictions generates network output
Noise image be real noise image probability.
In the optimal technical scheme of above scheme, it includes noise generation module and noise superposition that the noise, which generates network,
Module；
The noise generation module is configured to generate noise according to the low noise image sample and random noise and incite somebody to action
The noise is exported to the noise laminating module；
The noise laminating module is configured to the noise that the noise generation module exports being superimposed to the low noise sound spectrogram
Decent obtains the noise image.
In the optimal technical scheme of above scheme, the corresponding data matrix of the low noise image sample and described random
The corresponding data matrix size of noise is identical.
In the optimal technical scheme of above scheme, the system also includes network training device, the network training dress
It sets and is configured to according to the noise image sample and using machine learning algorithm simultaneously to generations confrontation network and described
Image denoising network carries out network training.
In the optimal technical scheme of above scheme, the generation confrontation network and described image denoising network are nerves
Network.
The second aspect of the present invention additionally provides a kind of image based on the removal system of picture noise described in above scheme
Noise remove method, the method includes being carried out at denoising using the image denoising network after completion network training to target image
Reason obtains low noise image；
The method also includes: using the network training device simultaneously to the generation confrontation network and figure in the system
As denoising network carries out network training.
In the optimal technical scheme of above scheme, " using the network training device simultaneously to the life in the system
At confrontation network and image denoising network carry out network training " the step of include:
According to preset noise image sample with preset objective function and using machine learning algorithm simultaneously to described
It generates confrontation network and described image denoising network carries out network training；
Wherein, the objective function includes that the corresponding first object function of the generation confrontation network and described image are gone
Corresponding second objective function of network of making an uproar.
In the optimal technical scheme of above scheme, the first object function is shown below:
Wherein, V (D, G) indicates that the first object function, the G and D respectively indicate in the generation confrontation network
Noise generation module and noise discrimination module, the pdataIt (x) is noise image sample set, the pzIt (z) is random noise letter
Number, the G (z) indicate the noise image that the noise generation module G is generated according to z-th of random noise, D (G (the z)) table
Show that the noise judgment module D predicts that the noise image G (z) is the probability of real noise image, described in D (x) expression
Noise judgment module D predicts that x-th of noise image sample is the probability of real noise image.
In the optimal technical scheme of above scheme, second objective function is minimum flat method error rule function.
The third aspect of the present invention additionally provides a kind of picture noise eliminating equipment, including control system and above scheme institute
The picture noise removal system stated, the control system include processor and storage equipment；
The storage equipment is suitable for storing a plurality of program, and described program is suitable for being loaded by the processor to execute above-mentioned side
Image noise elimination method described in case.
Compared with the immediate prior art, above-mentioned technical proposal is at least had the following beneficial effects:
Picture noise provided by the invention removes system, including generates confrontation network and image denoising network, generates confrontation
Network includes that noise generates network and noise differentiation network, and image denoising network can carry out denoising to target image and obtain
Low noise image, noise, which generates network, can be generated the forgery noise image close with real noise image, and noise differentiates network
It can differentiate that noise generates network and generates whether forgery noise is true noise image；System of the invention is not necessarily to image
Noise profile carry out it is any it is assumed that and without the use of pairing have noise and muting training data, can be adapted for various
The noise of type, and it is more economical.
Detailed description of the invention
Fig. 1 is the key step schematic diagram of the image noise elimination method of an embodiment of the present invention；
Fig. 2 (a) is the schematic diagram of the low noise image of an embodiment of the present invention；
Fig. 2 (b) is the schematic diagram of the analogue noise of an embodiment of the present invention；
Fig. 2 (c) is the schematic diagram of the forgery noise image of an embodiment of the present invention；
Fig. 2 (d) is the schematic diagram of the destination noise image of an embodiment of the present invention；
Fig. 2 (e) be an embodiment of the present invention noise reduction after image schematic diagram；
Fig. 3 is that the picture noise of an embodiment of the present invention removes the primary structure schematic diagram of system；
Fig. 4 is that the picture noise of an embodiment of the present invention removes the flow diagram that system removes noise.
Specific embodiment
In order to make the object, technical scheme and advantages of the embodiment of the invention clearer, below in conjunction with the embodiment of the present invention
In attached drawing, technical scheme in the embodiment of the invention is clearly and completely described, it is clear that described embodiment is
A part of the embodiment of the present invention, instead of all the embodiments.Based on the embodiments of the present invention, those of ordinary skill in the art
Every other embodiment obtained without making creative work, shall fall within the protection scope of the present invention.
The preferred embodiment of the present invention described with reference to the accompanying drawings.It will be apparent to a skilled person that this
A little embodiments are used only for explaining technical principle of the invention, it is not intended that limit the scope of the invention.
Refering to attached drawing 1, Fig. 1 schematically illustrates the key step of image noise elimination method in the present embodiment.Such as Fig. 1
Shown, image noise elimination method mainly includes the following steps: in the present embodiment
Step S101: low noise image and random noise corresponding with low noise image are obtained.
In practical applications, it is assumed that noise profile situation it is in real data and unreasonable, different location in image
Noise is likely to very strong correlation, and the noise profile situation of different images is different, cannot be reflected to the hypothesis of noise
True situation；Pairing has that noise and muting image data are difficult to obtain and procurement cost is higher.
Picture noise can be removed by trained neural fusion in advance, specifically, by taking CT image noise reduction as an example,
The low noise image and random noise corresponding with the image of available Low emissivity metering, wherein low noise image be with
The image that Low emissivity metering irradiation photo generates, noise in image data are lower, and random noise can be gaussian random noise, high
This random noise is a 2D matrix, the corresponding data matrix of low noise image data matrix size corresponding with random noise
It is identical.
Step S102: noise generates network and generates simulation according to low noise image and the corresponding random noise of low noise image
Noise.
Specifically, it is relevant in the value of different location due to true noise, and is with the signal value of image itself
It is relevant, and in order to retain the macrostructure of image script, network can be generated by noise and generate analogue noise, analogue noise
Only image is adjusted in part, thus avoid generate noise image and noise-free picture be it is corresponding, obtain more true
Real noise data.Wherein, noise generates network and can be according to preset noise image sample and preset objective function simultaneously
Using the trained neural network of machine learning algorithm, noise, which generates network, can be UNet network.
Low noise image and the corresponding random noise input noise of low noise image are generated into network, it is defeated that noise generates network
Analogue noise out, wherein the corresponding data matrix of analogue noise data matrix size corresponding with low noise image is identical, and mould
Quasi- noise is also a 2D matrix.
Step S103: analogue noise and low noise image are overlapped, and obtain forging noise image.
By the superposition of analogue noise and low noise image, the macrostructure of image script can be retained, noise is only in office
Portion is adjusted image, due to the problem of being subsequently generated the mode collapsing for fighting network, if directly generating noise
Image then cannot be guaranteed corresponding relationship.
Step S104: will forge noise image and low noise image input noise differentiates that network, noise differentiate that network differentiates
Input picture is the probability of real noise image.
Noise differentiates that network is a Classification Neural, classifies to two kinds of image, and noise generates network
The forgery noise image of generation can be close to true noise image, and noise differentiates that network can identify as much as possible
The noise image of forgery can make noise generate the noise image of network generation more by the training of multiple images sample data
Close to true noise image, differentiate that network can more accurately identify true image.
Wherein, noise differentiates that network can be according to preset noise image sample and preset first object function and benefit
With the trained neural network of machine learning algorithm, noise differentiates that network can be ResNet network.
Specifically, first object function can be as shown in formula (1):
Wherein, V (D, G) indicates that first object function, the G and D respectively indicate the noise generated in confrontation network and generate
Network and noise differentiate network, pdataIt (x) is noise image sample set, pzIt (z) is random noise function, G (z) indicates that noise is raw
At the noise image that network G is generated according to z-th of random noise, D (G (z)) indicates that noise judges that network D predicts noise image G
(z) be real noise image probability, the D (x) indicates that noise judgment module D predicts that x-th of noise image sample is true
The probability of noise image.
Step S105: image denoising network carries out denoising to target image, obtains low noise image.
In practical applications, noise being differentiated to, network is determined as really forging noise image input picture denoising net
Network, image denoising network according to preset noise image sample and preset second objective function and can utilize machine learning
The neural network of algorithm training, wherein the second objective function can be minimum flat method error rule function.
As shown in Fig. 2 (a)-(e), Fig. 2 (a)-(e) schematically illustrates low noise image in the present embodiment, simulation is made an uproar
The schematic diagram of image after sound, forgery noise image, destination noise image and noise reduction, wherein Fig. 2 (a) is low noise image, figure
2 (b) be analogue noise, and Fig. 2 (c) is to forge noise image, and Fig. 2 (d) is destination noise image, and Fig. 2 (e) is image after noise reduction.
It can be adapted for various types of noises by means of the present invention, be not necessarily to paired data, it is applied widely, and
More economically.
Although each step is described in the way of above-mentioned precedence in above-described embodiment, this field
Technical staff is appreciated that the effect in order to realize the present embodiment, executes between different steps not necessarily in such order,
It (parallel) execution simultaneously or can be executed with reverse order, these simple variations all protection scope of the present invention it
It is interior.
Based on above method embodiment, the present invention also provides a kind of picture noises to remove system.With reference to the accompanying drawing, right
Picture noise removal system is illustrated.
Refering to attached drawing 3, Fig. 3 schematically illustrates the primary structure of picture noise removal system in the present embodiment.Such as Fig. 3
Shown, it includes generating confrontation network and image denoising network 1 that the present embodiment picture noise, which removes system, generates confrontation network and includes
Noise generates network 2 and noise differentiates network 3.Refering to attached drawing 4, Fig. 4 schematically illustrates picture noise in the present embodiment and goes
Except the process of system removal noise.
Image denoising network 1 is configured to obtain low noise image to target image progress denoising, or in system pair
The output result for generating network 2 to noise when generating confrontation network and the progress network training of image denoising network 1 carries out at denoising
Reason obtains low noise image sample；
Noise generates network 2 and is configured to generate noise image according to low noise image sample and random noise and by noise
Image, which is exported respectively to image denoising network 1 and noise, differentiates network 3；
Noise differentiates that network 3 is configured to generate the noise that network 2 exports according to preset noise image sample predictions noise
Image is the probability of real noise image.
In the optimal technical scheme of above scheme, it includes noise generation module and noise superposition mould that noise, which generates network 2,
Block；
Noise generation module is configured to generate noise with random noise according to low noise image sample and exports noise
To noise laminating module；
Noise laminating module, which is configured to the noise that noise generation module exports being superimposed to low noise image sample, is made an uproar
Acoustic image.
In the optimal technical scheme of above scheme, the corresponding data matrix of low noise image sample is corresponding with random noise
Data matrix size it is identical.
In the optimal technical scheme of above scheme, system further includes network training device, and network training device is configured to
Network is carried out to generation confrontation network and image denoising network 1 simultaneously according to noise image sample and using machine learning algorithm
Training.
In the optimal technical scheme of above scheme, generates confrontation network and image denoising network 1 is neural network.
Based on the above system embodiment, the present invention also provides a kind of picture noise eliminating equipment, picture noise removals
Equipment may include control system and above-mentioned picture noise removal system, and control system may include that processor and storage are set
It is standby；
It stores equipment and is suitable for storing a plurality of program, program is suitable for being loaded by processor to execute above-mentioned picture noise removal
Method.
Person of ordinary skill in the field can be understood that, for convenience of description and succinctly, system of the present invention
The specific work process and related description of system, apparatus embodiments, can refer to corresponding processes in the foregoing method embodiment, and
With above method beneficial effect having the same, details are not described herein.
Those skilled in the art should be able to recognize that, side described in conjunction with the examples disclosed in the embodiments of the present disclosure
Method step, can be realized with electronic hardware, computer software, or a combination of the two, in order to clearly demonstrate electronic hardware and
The interchangeability of software generally describes each exemplary composition and step according to function in the above description.These
Function is executed actually with electronic hardware or software mode, specific application and design constraint depending on technical solution.
Those skilled in the art can use different methods to achieve the described function each specific application, but this reality
Now it should not be considered as beyond the scope of the present invention.
It should be noted that description and claims of this specification and term " first " in above-mentioned attached drawing, "
Two " etc. be to be used to distinguish similar objects, rather than be used to describe or indicate specific sequence or precedence.It should be understood that this
The data that sample uses can be interchanged in appropriate circumstances, so that the embodiment of the present invention described herein can be in addition at this
In illustrate or description those of other than sequence implement.
So far, it has been combined preferred embodiment shown in the drawings and describes technical solution of the present invention, still, this field
Technical staff is it is easily understood that protection scope of the present invention is expressly not limited to these specific embodiments.Without departing from this
Under the premise of the principle of invention, those skilled in the art can make equivalent change or replacement to the relevant technologies feature, these
Technical solution after change or replacement will fall within the scope of protection of the present invention.
Claims (10)
1. a kind of picture noise removes system, which is characterized in that including generating confrontation network and image denoising network, the generation
Confrontation network includes that noise generates network and noise differentiation network；
Described image denoising network configuration obtains low noise image to carry out denoising to target image, or in the system
The output knot of network is generated when carrying out network training to the generation confrontation network and described image denoising network to the noise
Fruit carries out denoising and obtains low noise image sample；
It is to generate noise image according to the low noise image sample and random noise and will that the noise, which generates network configuration,
The noise image exports respectively to described image and denoises network and noise differentiation network；
The noise differentiates that network configuration is that the noise according to preset noise image sample predictions generates making an uproar for network output
Acoustic image is the probability of real noise image.
2. picture noise according to claim 1 removes system, which is characterized in that it includes noise that the noise, which generates network,
Generation module and noise laminating module；
The noise generation module is configured to generate noise according to the low noise image sample and random noise and will be described
Noise is exported to the noise laminating module；
The noise laminating module is configured to the noise that the noise generation module exports being superimposed to the low noise image sample
Originally the noise image is obtained.
3. picture noise according to claim 2 removes system, which is characterized in that the low noise image sample is corresponding
Data matrix data matrix size corresponding with the random noise is identical.
4. picture noise according to claim 2 removes system, which is characterized in that the system also includes network training dresses
It sets, the network training device is configured to according to the noise image sample and using machine learning algorithm simultaneously to the life
Network training is carried out at confrontation network and described image denoising network.
5. picture noise according to any one of claim 1 to 4 removes system, which is characterized in that the generation confrontation
Network and described image denoising network are neural networks.
6. a kind of image noise elimination method based on the removal system of picture noise described in any one of claims 1 to 5,
It is characterized in that, the method includes carrying out denoising to target image using the image denoising network after completion network training to obtain
To low noise image；
The method also includes: network is fought to the generation in the system simultaneously using the network training device and image is gone
Network of making an uproar carries out network training.
7. image noise elimination method according to claim 6, which is characterized in that " same using the network training device
When in the system generation confrontation network and image denoising network carry out network training " the step of include:
According to preset noise image sample and preset objective function and using machine learning algorithm simultaneously to the generation
It fights network and described image denoising network carries out network training；
Wherein, the objective function includes the corresponding first object function of the generation confrontation network and described image denoising net
Corresponding second objective function of network.
8. image noise elimination method according to claim 7, which is characterized in that the first object function such as following formula institute
Show:
Wherein, V (D, G) indicates that the first object function, the G and D respectively indicate the noise generated in confrontation network
Generation module and noise discrimination module, the pdataIt (x) is noise image sample set, the pzIt (z) is random noise function, institute
Stating G (z) indicates the noise image that the noise generation module G is generated according to z-th of random noise, and the D (G (z)) indicates institute
It states noise judgment module D and predicts that the noise image G (z) is the probability of real noise image, the D (x) indicates the noise
Judgment module D predicts that x-th of noise image sample is the probability of real noise image.
9. image noise elimination method according to claim 8, which is characterized in that second objective function is minimum flat
Method error rule function.
10. a kind of picture noise eliminating equipment, which is characterized in that including described in control system and any one of claims 1 to 5
Picture noise remove system, the control system include processor and storage equipment；
The storage equipment is suitable for storing a plurality of program, and described program, which is suitable for being loaded by the processor, requires 6 with perform claim
To image noise elimination method described in any one of 9.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811459206.5A CN109636746B (en) | 2018-11-30 | 2018-11-30 | Image noise removing system, method and equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811459206.5A CN109636746B (en) | 2018-11-30 | 2018-11-30 | Image noise removing system, method and equipment |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109636746A true CN109636746A (en) | 2019-04-16 |
| CN109636746B CN109636746B (en) | 2020-09-08 |
Family
ID=66070590
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811459206.5A Active CN109636746B (en) | 2018-11-30 | 2018-11-30 | Image noise removing system, method and equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109636746B (en) |
Cited By (4)
| 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 |
| CN110517204A (en) * | 2019-08-30 | 2019-11-29 | 京东方科技集团股份有限公司 | A kind of noise cancellation method and device, detector of X-ray detector |
| CN111583144A (en) * | 2020-04-30 | 2020-08-25 | 深圳市商汤智能传感科技有限公司 | Image noise reduction method and device, electronic equipment and storage medium |
| CN113538223A (en) * | 2021-07-28 | 2021-10-22 | 维沃移动通信有限公司 | Noise image generation method, noise image generation device, electronic device, and storage medium |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106408522A (en) * | 2016-06-27 | 2017-02-15 | 深圳市未来媒体技术研究院 | Image de-noising method based on convolution pair neural network |
| 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 |
| WO2017223560A1 (en) * | 2016-06-24 | 2017-12-28 | Rensselaer Polytechnic Institute | Tomographic image reconstruction via machine learning |
| WO2018053340A1 (en) * | 2016-09-15 | 2018-03-22 | Twitter, Inc. | Super resolution using a generative adversarial network |
| CN108198154A (en) * | 2018-03-19 | 2018-06-22 | 中山大学 | Image de-noising method, device, equipment and storage medium |
| CN108416752A (en) * | 2018-03-12 | 2018-08-17 | 中山大学 | A method of image is carried out based on production confrontation network and removes motion blur |
| CN108492258A (en) * | 2018-01-17 | 2018-09-04 | 天津大学 | A kind of radar image denoising method based on generation confrontation network |
| CN108765319A (en) * | 2018-05-09 | 2018-11-06 | 大连理工大学 | A kind of image de-noising method based on generation confrontation network |
-
2018
- 2018-11-30 CN CN201811459206.5A patent/CN109636746B/en active Active
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2017223560A1 (en) * | 2016-06-24 | 2017-12-28 | Rensselaer Polytechnic Institute | Tomographic image reconstruction via machine learning |
| CN106408522A (en) * | 2016-06-27 | 2017-02-15 | 深圳市未来媒体技术研究院 | Image de-noising method based on convolution pair neural network |
| WO2018053340A1 (en) * | 2016-09-15 | 2018-03-22 | Twitter, Inc. | Super resolution using a generative adversarial network |
| 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 |
| CN108492258A (en) * | 2018-01-17 | 2018-09-04 | 天津大学 | A kind of radar image denoising method based on generation confrontation network |
| CN108416752A (en) * | 2018-03-12 | 2018-08-17 | 中山大学 | A method of image is carried out based on production confrontation network and removes motion blur |
| CN108198154A (en) * | 2018-03-19 | 2018-06-22 | 中山大学 | Image de-noising method, device, equipment and storage medium |
| CN108765319A (en) * | 2018-05-09 | 2018-11-06 | 大连理工大学 | A kind of image de-noising method based on generation confrontation network |
Non-Patent Citations (2)
| Title |
|---|
| QINGSONG YANG,ET AL: "Low-Dose CT Image Denoising Using a", 《IEEE TRANSACTIONS ON MEDICAL IMAGING》 * |
| 张清博 等: "基于改进生成对抗网络的水下激光图像后向散射光修复方法", 《基于改进生成对抗网络的水下激光图像后向》 * |
Cited By (8)
| 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 |
| 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 |
| CN110517204A (en) * | 2019-08-30 | 2019-11-29 | 京东方科技集团股份有限公司 | A kind of noise cancellation method and device, detector of X-ray detector |
| CN110517204B (en) * | 2019-08-30 | 2022-05-20 | 京东方科技集团股份有限公司 | Noise elimination method and device of X-ray detector and detector |
| CN111583144A (en) * | 2020-04-30 | 2020-08-25 | 深圳市商汤智能传感科技有限公司 | Image noise reduction method and device, electronic equipment and storage medium |
| CN111583144B (en) * | 2020-04-30 | 2023-08-25 | 深圳市商汤智能传感科技有限公司 | Image noise reduction method and device, electronic equipment and storage medium |
| CN113538223A (en) * | 2021-07-28 | 2021-10-22 | 维沃移动通信有限公司 | Noise image generation method, noise image generation device, electronic device, and storage medium |
| WO2023005818A1 (en) * | 2021-07-28 | 2023-02-02 | 维沃移动通信有限公司 | Noise image generation method and apparatus, electronic device, and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109636746B (en) | 2020-09-08 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109636746A (en) | Picture noise removes system, method and apparatus | |
| US11093780B2 (en) | Device and method to generate image using image learning model | |
| CN109858461B (en) | Method, device, equipment and storage medium for counting dense population | |
| CN108875732B (en) | Model training and instance segmentation method, device and system and storage medium | |
| KR101803081B1 (en) | Robot for store management | |
| CN111340008B (en) | Method and system for generation of counterpatch, training of detection model and defense of counterpatch | |
| CN106485215B (en) | Face shielding detection method based on deep convolutional neural network | |
| JP6772023B2 (en) | Method and system of collective interaction by user state detection | |
| CN109214973B (en) | Method for generating countermeasure security carrier aiming at steganalysis neural network | |
| CN108182384B (en) | Face feature point positioning method and device | |
| CN109214337A (en) | A kind of Demographics' method, apparatus, equipment and computer readable storage medium | |
| CN107680593A (en) | The sound enhancement method and device of a kind of smart machine | |
| Ramsey et al. | Non-Gaussian methods and high-pass filters in the estimation of effective connections | |
| CN108664880A (en) | Activity test method and equipment | |
| US20210042497A1 (en) | Visual fatigue recognition method, visual fatigue recognition device, virtual reality apparatus and storage medium | |
| CN108875469A (en) | In vivo detection and identity authentication method, device and computer storage medium | |
| CN110149476A (en) | A kind of time-lapse photography method, apparatus, system and terminal device | |
| CN111832517A (en) | Low-definition face key point detection method based on gated convolution | |
| CN113570689B (en) | Portrait cartoon method, device, medium and computing equipment | |
| CN112509154B (en) | Training method of image generation model, image generation method and device | |
| CN111476741B (en) | Image denoising method, image denoising device, electronic equipment and computer readable medium | |
| CN109101858A (en) | Action identification method and device | |
| CN112149500A (en) | Partially-shielded face recognition small sample learning method | |
| CN111144497A (en) | Image significance prediction method under multitask depth network based on aesthetic analysis | |
| CN109886402A (en) | Deep learning model training method, device, computer equipment and storage medium |
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 |