CN110415187A - Image processing method and image processing system - Google Patents
Image processing method and image processing system Download PDFInfo
- Publication number
- CN110415187A CN110415187A CN201910610059.5A CN201910610059A CN110415187A CN 110415187 A CN110415187 A CN 110415187A CN 201910610059 A CN201910610059 A CN 201910610059A CN 110415187 A CN110415187 A CN 110415187A
- Authority
- CN
- China
- Prior art keywords
- image
- convolutional neural
- size
- neural networks
- image acquisition
- 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
- 238000012545 processing Methods 0.000 title claims abstract description 22
- 238000003672 processing method Methods 0.000 title claims abstract description 17
- 238000012546 transfer Methods 0.000 claims abstract description 89
- 238000013527 convolutional neural network Methods 0.000 claims abstract description 66
- 230000007704 transition Effects 0.000 claims abstract description 52
- 238000000034 method Methods 0.000 claims abstract description 36
- 230000008569 process Effects 0.000 claims abstract description 30
- 238000013528 artificial neural network Methods 0.000 claims abstract description 15
- 239000011159 matrix material Substances 0.000 claims abstract description 11
- 230000002708 enhancing effect Effects 0.000 claims description 9
- 230000001537 neural effect Effects 0.000 claims 1
- 230000007812 deficiency Effects 0.000 description 3
- 230000008859 change Effects 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 230000035772 mutation Effects 0.000 description 2
- 241000276489 Merlangius merlangus Species 0.000 description 1
- 230000003321 amplification Effects 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000002490 cerebral effect Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000010365 information processing Effects 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 210000005036 nerve Anatomy 0.000 description 1
- 238000003199 nucleic acid amplification method Methods 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/50—Image enhancement or restoration using two or more images, e.g. averaging or subtraction
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/90—Dynamic range modification of images or parts thereof
-
- 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
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Image Processing (AREA)
Abstract
The present invention provides a kind of image processing method and image processing system.The image processing method includes the following steps: to receive original image, and reduces the size of original image, obtains First Transition image；First expansion convolutional neural networks are provided, expansion process of convolution is carried out to First Transition image using the first expansion convolutional neural networks, to enhance the image quality of First Transition image, obtains the second transfer image acquisition；The size for amplifying the second transfer image acquisition, obtains third transfer image acquisition, and the size of third transfer image acquisition and the size of original image are identical；Original image and third transfer image acquisition matrix are spliced, the 4th transfer image acquisition is obtained；Second expansion convolutional neural networks are provided, expansion process of convolution is carried out to the 4th transfer image acquisition using the second expansion convolutional neural networks, to restore the detailed information of the 4th transfer image acquisition, obtains target image.The present invention can reduce neural network computing amount in the case where not reducing image detail information, and avoiding that treated, abrupt local occurs in image.
Description
Technical field
The present invention relates to field of display technology more particularly to a kind of image processing methods and image processing system.
Background technique
With the development of science and technology, requirement of the people to display device is higher and higher, it is lighter, apparent, more bright-coloured to become
The developing direction and target of TV.In addition to improving TV resolution on hardware, increase the side such as dynamic range and gamut range
Formula can also believe image by the way that image processing system commander in a display device is arranged and coordinates display device various functions
Number optimize one by one, so as to bring better image quality to user on the basis of existing hardware,
Image enhancement technique is one kind of image processing techniques, it can significantly improve picture quality, so that picture material
More have a sense of hierarchy and subjective observation effect more meets people's demand.In actual life, often there are various problems in original image,
Such as: aperture is less than normal when taking pictures, and causes image partially dark；The contrast of scene is lower, and image emphasis is not protruded；It is exposed
Degree, causes image not normal, photo whiting etc..It can effectively be solved the above problems by image enhancement technique, promote display quality.
Common image enhancement technique includes: saturation degree enhancing and contrast enhancing, is enhanced compared to saturation degree, contrast
It is higher to enhance the attention rate being subject to.Contrast enhancing is the gray-scale distribution by adjusting image, increases the distribution model of image gray-scale level
It encloses, to improve the contrast of image in whole or in part, improves visual effect.
Neural network framework is a kind of mathematics of structure progress information processing that application couples similar to cerebral nerve cynapse
Model.Its sharpest edges is the mechanism that can be used as an arbitrary function and approach, from the data " study " observed.In recent years
Come, neural network framework is increasingly being used on image processing algorithm, and convolutional neural networks are typically for example expanded.
The feature that convolutional neural networks are capable of reason expansion convolution is expanded, while expanding neural network view field, guarantor
Less operation times are held, realize the neural network output of original image pixel precision, but for large-size images, expand convolutional Neural
Network there are problems that causing image local to be mutated because of view field deficiency, and the prior art is forced to deepen network deep for the problem
Degree, leads to network capacity redundancy, and the complexity of the network architecture increases.
Summary of the invention
The purpose of the present invention is to provide a kind of image processing methods, can reduce neural network computing amount, avoid enhancing
There is abrupt local in image afterwards, and does not reduce the detailed information of enhanced image.
The object of the invention is also to provide a kind of image processing systems, can reduce neural network computing amount, avoid increasing
There is abrupt local in image after strong, and does not reduce the detailed information of enhanced image.
To achieve the above object, the present invention provides a kind of image processing method, includes the following steps:
Step S1, original image is received, and reduces the size of original image, obtains First Transition image；
Step S2, the first expansion convolutional neural networks are provided, using the first expansion convolutional neural networks to described the
One transfer image acquisition carries out expansion process of convolution, to enhance the image quality of First Transition image, obtains the second transfer image acquisition；
Step S3, the size for amplifying second transfer image acquisition, obtains third transfer image acquisition, the third transfer image acquisition
Size is identical as the size of original image；
Step S4, the original image and the third transfer image acquisition matrix are spliced, obtains the 4th transfer image acquisition；
Step S5, the second expansion convolutional neural networks are provided, using the second expansion convolutional neural networks to described the
Four transfer image acquisitions carry out expansion process of convolution, to restore the detailed information of the 4th transfer image acquisition, obtain target image.
The size of the First Transition image is less than or equal to 256 pixels × 256 pixels.
The first expansion convolutional neural networks carry out at least five layers expansion process of convolution to the First Transition image.
The second expansion convolutional neural networks carry out at least two layers expansion process of convolution to the 4th transfer image acquisition.
The size of original image is reduced in the step S1 by bilinear interpolation algorithm, passes through two-wire in the step S3
Property interpolation algorithm amplify the second transfer image acquisition size.
The present invention also provides a kind of image processing systems, including input unit, the diminution list being connected with the input unit
Member, first to be connected with the reducing unit expansion convolutional neural networks expand what convolutional neural networks were connected with described first
Amplifying unit, the concatenation unit being connected with the amplifying unit and input unit, the second expansion being connected with the concatenation unit
Convolutional neural networks and the output unit being connected with the second expansion convolutional neural networks；
Original image is supplied to reducing unit and concatenation unit for receiving original image by the input unit；
The reducing unit is used to reduce the size of original image, obtains First Transition image；
The first expansion convolutional neural networks are for carrying out expansion process of convolution to the First Transition image, with enhancing
The image quality of First Transition image obtains the second transfer image acquisition；
The amplifying unit is used to amplify the size of the second transfer image acquisition, obtains third transfer image acquisition, the third transition
The size of image and the size of original image are identical；
The concatenation unit is used to carry out matrix splicing to the original image and third transfer image acquisition, obtains the 4th transition
Image；
The second expansion convolutional neural networks are for carrying out expansion process of convolution to the 4th transfer image acquisition, to restore
The detailed information of 4th transfer image acquisition, obtains target image；
The output unit is for exporting target image.
The size of the First Transition image is less than or equal to 256 pixels × 256 pixels.
The first expansion convolutional neural networks carry out at least five layers expansion process of convolution to the First Transition image.
The second expansion convolutional neural networks carry out at least two layers expansion process of convolution to the 4th transfer image acquisition.
The reducing unit reduces the size of original image by bilinear interpolation algorithm, and the amplifying unit passes through two-wire
Property interpolation algorithm amplify the second transfer image acquisition size.
Beneficial effects of the present invention: the present invention provides a kind of image processing method, includes the following steps: to receive original graph
Picture, and the size of original image is reduced, obtain First Transition image；First expansion convolutional neural networks are provided, utilize described the
One expansion convolutional neural networks carry out expansion process of convolution to the First Transition image, to enhance the picture of First Transition image
Matter obtains the second transfer image acquisition；The size for amplifying second transfer image acquisition obtains third transfer image acquisition, the third transition
The size of image and the size of original image are identical；The original image and the third transfer image acquisition matrix are spliced, obtained
4th transfer image acquisition；The second expansion convolutional neural networks are provided, using the second expansion convolutional neural networks to the described 4th
Transfer image acquisition carries out expansion process of convolution, to restore the detailed information of the 4th transfer image acquisition, obtains target image, can not subtract
In the case where few image detail information, neural network computing amount is reduced, avoiding that treated, abrupt local occurs in image.The present invention
A kind of image processing system is also provided, neural network computing amount can be reduced in the case where not reducing image detail information, keep away
Exempting from treated, there is abrupt local in image.
Detailed description of the invention
For further understanding of the features and technical contents of the present invention, it please refers to below in connection with of the invention detailed
Illustrate and attached drawing, however, the drawings only provide reference and explanation, is not intended to limit the present invention.
In attached drawing,
Fig. 1 is the flow chart of image processing method of the invention；
Fig. 2 is the image change schematic diagram of image processing method of the invention；
Fig. 3 is the schematic diagram of image processing system of the invention.
Specific embodiment
Further to illustrate technological means and its effect adopted by the present invention, below in conjunction with preferred implementation of the invention
Example and its attached drawing are described in detail.
Fig. 1 and Fig. 2 is please referred to, the present invention provides a kind of image processing method, includes the following steps:
Step S1, original image 1 is received, and reduces the size of original image 1, obtains First Transition image 2；
Specifically, the size of the First Transition image 2 is less than or equal to 256 pixels × 256 pixels, it is preferable that described
The size of First Transition image 2 is equal to 256 pixels × 256 pixels.
Specifically, the size of original image 1 is reduced in the step S1 by bilinear interpolation algorithm.
Step S2, the first expansion convolutional neural networks 30 are provided, first expansion, 30 pairs of institutes of convolutional neural networks are utilized
It states First Transition image 2 and carries out expansion process of convolution, to enhance the image quality of First Transition image, obtain the second transfer image acquisition 3.
Specifically, the first expansion convolutional neural networks 30 carry out at least five layers expansion to the First Transition image 2
Process of convolution, it is preferable that the first expansion convolutional neural networks 30 carry out five layers of expansion convolution to the First Transition image 2
Processing.
Step S3, the size for amplifying second transfer image acquisition 3, obtains third transfer image acquisition 4, the third transfer image acquisition
4 size is identical as the size of original image 1.
Specifically, amplify the size of the second transfer image acquisition 3 in the step S3 by bilinear interpolation algorithm.
Step S4, the original image 1 is spliced with 4 matrix of third transfer image acquisition, obtains the 4th transfer image acquisition 5.
Step S5, the second expansion convolutional neural networks 60 are provided, second expansion, 60 pairs of institutes of convolutional neural networks are utilized
It states the 4th transfer image acquisition 5 and carries out expansion process of convolution, to restore the detailed information of the 4th transfer image acquisition 5, obtain target image 6.
Specifically, the second expansion convolutional neural networks 60 carry out at least two layers expansion to the 4th transfer image acquisition 5
Process of convolution, it is preferable that the second expansion convolutional neural networks 60 carry out two layers of expansion convolution to the 4th transfer image acquisition 5
Processing.
It should be noted that the present invention is by dwindling into fixed-size First Transition image 2 for original image 1, in ruler
Image enhancement is carried out on very little lesser First Transition image 2, to guarantee the first expansion convolutional neural networks 30 during enhancing
There is sufficient view field to cover whole image, to realize when carrying out image enhancement to various sizes of original image always
Sufficient view field is all had, picture abrupt local caused by avoiding because of view field deficiency, and greatly reduce the first expansion
The operand of convolutional neural networks 30, reduces network cost.
Further, the second transfer image acquisition 3 that the present invention expands the enhanced small size of convolutional neural networks 30 for first
It amplifies, obtains the third transfer image acquisition 4 of size identical as original image, and connect third transfer image acquisition 4 by transition
Matrix splicing is carried out with original image 1, restores the size of image, is finally restored again by the second expansion convolutional neural networks 30
Image detail information obtains enhanced target image 6, and target image 6 is identical as the size of original image 1, thus of the invention
Neural network computing amount can be reduced in the case where not reducing image detail information, avoiding that treated, part occurs in image
Mutation.
Fig. 3 and Fig. 2 is please referred to, the present invention also provides a kind of image processing systems, including input unit 10 and the input
The connected reducing unit 20 of unit 10, first to be connected with the reducing unit 20 expand convolutional neural networks 30 and described the
The one splicing list for expanding the connected amplifying unit 40 of convolutional neural networks 30, being connected with the amplifying unit 40 and input unit 10
Member 50, second to be connected with the concatenation unit 50 expand convolutional neural networks 60 and expand 60 phase of convolutional neural networks with second
Output unit 70 even；
The input unit 10 is supplied to reducing unit 20 and splicing list for receiving original image 1, and by original image 1
Member 50；
The reducing unit 20 is used to reduce the size of original image 1, obtains First Transition image 2；
The first expansion convolutional neural networks 30 are used to carry out expansion process of convolution to the First Transition image 2, with
The image quality for enhancing First Transition image, obtains the second transfer image acquisition 3；
The amplifying unit 40 is used to amplify the size of the second transfer image acquisition 3, obtains third transfer image acquisition 4, the third
The size of transfer image acquisition 4 is identical as the size of original image 1；
The concatenation unit 50 is used to carry out matrix splicing to the original image 1 and third transfer image acquisition 4, obtains the 4th
Transfer image acquisition 5；
The second expansion convolutional neural networks 60 are used to carry out expansion process of convolution to the 4th transfer image acquisition 5, with
The detailed information for restoring the 4th transfer image acquisition 5 obtains target image 6；
The output unit 70 is for exporting target image 6.
Specifically, the size of the First Transition image 2 is less than or equal to 256 pixels × 256 pixels, it is preferable that described
The size of First Transition image 2 is equal to 256 pixels × 256 pixels
Specifically, the first expansion convolutional neural networks 30 carry out at least five layers expansion to the First Transition image 2
Process of convolution, it is preferable that the first expansion convolutional neural networks 30 carry out five layers of expansion convolution to the First Transition image 2
Processing
Specifically, the second expansion convolutional neural networks 60 carry out at least two layers expansion to the 4th transfer image acquisition 5
Process of convolution, it is preferable that the second expansion convolutional neural networks 60 carry out two layers of expansion convolution to the 4th transfer image acquisition 5
Processing.
Specifically, the reducing unit 20 reduces the size of original image 1 by bilinear interpolation algorithm, and the amplification is single
Member 40 amplifies the size of the second transfer image acquisition 3 by bilinear interpolation algorithm.
It should be noted that the present invention is by dwindling into fixed-size First Transition image 2 for original image 1, in ruler
Image enhancement is carried out on very little lesser First Transition image 2, to guarantee the first expansion convolutional neural networks 30 during enhancing
There is sufficient view field to cover whole image, to realize when carrying out image enhancement to various sizes of original image always
Sufficient view field is all had, picture abrupt local caused by avoiding because of view field deficiency, and greatly reduce the first expansion
The operand of convolutional neural networks 30, reduces network cost.
Further, the second transfer image acquisition 3 that the present invention expands the enhanced small size of convolutional neural networks 30 for first
It amplifies, obtains the third transfer image acquisition 4 of size identical as original image, and connect third transfer image acquisition 4 by transition
Matrix splicing is carried out with original image 1, restores the size of image, is finally restored again by the second expansion convolutional neural networks 30
Image detail information obtains enhanced target image 6, and target image 6 is identical as the size of original image 1, thus of the invention
Neural network computing amount can be reduced in the case where not reducing image detail information, avoiding that treated, part occurs in image
Mutation.
In conclusion the present invention provides a kind of image processing method, include the following steps: to receive original image, and reduce
The size of original image obtains First Transition image；First expansion convolutional neural networks are provided, the first expansion convolution is utilized
Neural network carries out expansion process of convolution to the First Transition image, to enhance the image quality of First Transition image, obtains second
Transfer image acquisition；The size for amplifying second transfer image acquisition, obtains third transfer image acquisition, the size of the third transfer image acquisition with
The size of original image is identical；The original image and the third transfer image acquisition matrix are spliced, the 4th transfer image acquisition is obtained；
Second expansion convolutional neural networks are provided, the 4th transfer image acquisition are carried out using the second expansion convolutional neural networks swollen
Swollen process of convolution obtains target image, can not reduce image detail information to restore the detailed information of the 4th transfer image acquisition
In the case where, neural network computing amount is reduced, avoiding that treated, abrupt local occurs in image.The present invention also provides a kind of images
Processing system can reduce neural network computing amount in the case where not reducing image detail information, the image that avoids that treated
There is abrupt local.
The above for those of ordinary skill in the art can according to the technique and scheme of the present invention and technology
Other various corresponding changes and modifications are made in design, and all these change and modification all should belong to the claims in the present invention
Protection scope.
Claims (10)
1. a kind of image processing method, which comprises the steps of:
Step S1, original image (1) is received, and reduces the size of original image (1), obtain First Transition image (2)；
Step S2, the first expansion convolutional neural networks (30) is provided, using first expansion convolutional neural networks (30) to institute
It states First Transition image (2) and carries out expansion process of convolution, to enhance the image quality of First Transition image, obtain the second transfer image acquisition
(3)；
Step S3, the size for amplifying second transfer image acquisition (3), obtains third transfer image acquisition (4), the third transfer image acquisition
(4) size is identical as the size of original image (1)；
Step S4, the original image (1) and third transfer image acquisition (4) matrix are spliced, obtains the 4th transfer image acquisition
(5)；
Step S5, the second expansion convolutional neural networks (60) is provided, using second expansion convolutional neural networks (60) to institute
It states the 4th transfer image acquisition (5) and carries out expansion process of convolution, to restore the detailed information of the 4th transfer image acquisition (5), obtain target figure
As (6).
2. image processing method as described in claim 1, which is characterized in that the size of the First Transition image (2) is less than
Or it is equal to 256 pixels × 256 pixels.
3. image processing method as described in claim 1, which is characterized in that first expansion convolutional neural networks (30) is right
The First Transition image (2) carries out at least five layers expansion process of convolution.
4. image processing method as described in claim 1, which is characterized in that second expansion convolutional neural networks (60) is right
4th transfer image acquisition (5) carries out at least two layers expansion process of convolution.
5. image processing method as described in claim 1, which is characterized in that pass through bilinear interpolation algorithm in the step S1
The size of original image (1) is reduced, amplifies the ruler of the second transfer image acquisition (3) in the step S3 by bilinear interpolation algorithm
It is very little.
6. a kind of image processing system, which is characterized in that including input unit (10), the contracting being connected with the input unit (10)
Junior unit (20), first to be connected with the reducing unit (20) expand convolutional neural networks (30), roll up with first expansion
The splicing list for accumulating neural network (30) connected amplifying unit (40), being connected with the amplifying unit (40) and input unit (10)
First (50), second to be connected with the concatenation unit (50) expand convolutional neural networks (60) and expand convolutional Neural net with second
The connected output unit (70) of network (60)；
Original image (1) is supplied to reducing unit (20) and spelled by the input unit (10) for receiving original image (1)
Order member (50)；
The reducing unit (20) is used to reduce the size of original image (1), obtains First Transition image (2)；
First expansion convolutional neural networks (30) is used to carry out expansion process of convolution to the First Transition image (2), with
The image quality for enhancing First Transition image, obtains the second transfer image acquisition (3)；
The amplifying unit (40) is used to amplify the sizes of the second transfer image acquisition (3), obtains third transfer image acquisition (4), and described the
The size of three transfer image acquisitions (4) is identical as the size of original image (1)；
The concatenation unit (50) is used to carry out matrix splicing to the original image (1) and third transfer image acquisition (4), obtains the
Four transfer image acquisitions (5)；
Second expansion convolutional neural networks (60) is used to carry out expansion process of convolution to the 4th transfer image acquisition (5), with
The detailed information for restoring the 4th transfer image acquisition (5) obtains target image (6)；
The output unit (70) is for exporting target image (6).
7. image processing system as claimed in claim 6, which is characterized in that the size of the First Transition image (2) is less than
Or it is equal to 256 pixels × 256 pixels.
8. image processing system as claimed in claim 6, which is characterized in that first expansion convolutional neural networks (30) is right
The First Transition image (2) carries out at least five layers expansion process of convolution.
9. image processing system as claimed in claim 6, which is characterized in that second expansion convolutional neural networks (60) is right
4th transfer image acquisition (5) carries out at least two layers expansion process of convolution.
10. image processing system as claimed in claim 6, which is characterized in that the reducing unit (20) is inserted by bilinearity
Value-based algorithm reduces the size of original image (1), and the amplifying unit (40) amplifies the second transition figure by bilinear interpolation algorithm
As the size of (3).
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910610059.5A CN110415187B (en) | 2019-07-04 | 2019-07-04 | Image processing method and image processing system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910610059.5A CN110415187B (en) | 2019-07-04 | 2019-07-04 | Image processing method and image processing system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110415187A true CN110415187A (en) | 2019-11-05 |
| CN110415187B CN110415187B (en) | 2021-07-23 |
Family
ID=68360596
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910610059.5A Active CN110415187B (en) | 2019-07-04 | 2019-07-04 | Image processing method and image processing system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110415187B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2021179147A1 (en) * | 2020-03-09 | 2021-09-16 | 华为技术有限公司 | Image processing method and apparatus based on neural network |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108038893A (en) * | 2017-12-06 | 2018-05-15 | 深圳市未来媒体技术研究院 | A kind of generation method of 1,000,000,000 pixel videos based on Hybrid camera array |
| CN108376387A (en) * | 2018-01-04 | 2018-08-07 | 复旦大学 | Image deblurring method based on polymerization expansion convolutional network |
| US10068557B1 (en) * | 2017-08-23 | 2018-09-04 | Google Llc | Generating music with deep neural networks |
| CN108734211A (en) * | 2018-05-17 | 2018-11-02 | 腾讯科技（深圳）有限公司 | The method and apparatus of image procossing |
| CN108830809A (en) * | 2018-06-05 | 2018-11-16 | 陕西师范大学 | One kind is based on expansion convolved image denoising method |
| CN108876805A (en) * | 2018-06-20 | 2018-11-23 | 长安大学 | The end-to-end unsupervised scene of one kind can traffic areas cognition and understanding method |
| US20190050667A1 (en) * | 2017-03-10 | 2019-02-14 | TuSimple | System and method for occluding contour detection |
-
2019
- 2019-07-04 CN CN201910610059.5A patent/CN110415187B/en active Active
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20190050667A1 (en) * | 2017-03-10 | 2019-02-14 | TuSimple | System and method for occluding contour detection |
| US10068557B1 (en) * | 2017-08-23 | 2018-09-04 | Google Llc | Generating music with deep neural networks |
| CN108038893A (en) * | 2017-12-06 | 2018-05-15 | 深圳市未来媒体技术研究院 | A kind of generation method of 1,000,000,000 pixel videos based on Hybrid camera array |
| CN108376387A (en) * | 2018-01-04 | 2018-08-07 | 复旦大学 | Image deblurring method based on polymerization expansion convolutional network |
| CN108734211A (en) * | 2018-05-17 | 2018-11-02 | 腾讯科技（深圳）有限公司 | The method and apparatus of image procossing |
| CN108830809A (en) * | 2018-06-05 | 2018-11-16 | 陕西师范大学 | One kind is based on expansion convolved image denoising method |
| CN108876805A (en) * | 2018-06-20 | 2018-11-23 | 长安大学 | The end-to-end unsupervised scene of one kind can traffic areas cognition and understanding method |
Non-Patent Citations (4)
| Title |
|---|
| MUYI SUN等: "Accurate Gastric Cancer Segmentation in Digital Pathology Images Using Deformable Convolution and Multi-Scale Embedding Networks", 《IEEE ACCESS》 * |
| WEIQIAN LU等: "Salient object detection via multi-scale neural network", 《PROCEEDINGS OF THE 10TH INTERNATIONAL CONFERENCE ON INTERNET MULTIMEDIA COMPUTING AND SERVICE》 * |
| 吕培建 等: "基于上下文以及多尺度信息融合的目标检测算法", 《计算机科学》 * |
| 李宝奇: "基于全卷积神经网络的非对称并行语义分割模型", 《电子学报》 * |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2021179147A1 (en) * | 2020-03-09 | 2021-09-16 | 华为技术有限公司 | Image processing method and apparatus based on neural network |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110415187B (en) | 2021-07-23 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US10861133B1 (en) | Super-resolution video reconstruction method, device, apparatus and computer-readable storage medium | |
| CN108961186B (en) | Old film repairing and reproducing method based on deep learning | |
| CN101441764B (en) | MTFC remote sensing image restoration method | |
| CN105719611B (en) | The show uniformity method of adjustment and device of liquid crystal display | |
| US8837854B2 (en) | Image processing method for boundary resolution enhancement | |
| WO2023010754A1 (en) | Image processing method and apparatus, terminal device, and storage medium | |
| US10007970B2 (en) | Image up-sampling with relative edge growth rate priors | |
| CN105009168A (en) | Restoration filter generation device and method, image processing device and method, imaging device, program, and recording medium | |
| CN106920221A (en) | Take into account the exposure fusion method that Luminance Distribution and details are presented | |
| US8860806B2 (en) | Method, device, and system for performing color enhancement on whiteboard color image | |
| CA2554989A1 (en) | Super-resolution image processing | |
| CN113939845A (en) | Method, system and computer readable medium for improving image color quality | |
| WO2020118902A1 (en) | Image processing method and image processing system | |
| CN109300092A (en) | A kind of remote sensing images deblurring method based on dark | |
| CN115115516A (en) | Real-world video super-resolution algorithm based on Raw domain | |
| CN110415187A (en) | Image processing method and image processing system | |
| JPH08305846A (en) | Neuro filter, image area dividing method, and filter device | |
| CN108521535A (en) | A kind of Information hiding transmission method based on image blend processing | |
| CN104361559A (en) | Image magnifying method and system based on small image memory | |
| CN110390636A (en) | Unmanned plane is super, the simulation Zooming method of high definition picture or video data | |
| US9911178B2 (en) | System and method for content-adaptive super-resolution via cross-scale self-learning | |
| CN111009208B (en) | Optimization method of panel display, display panel and storage medium | |
| CN107590790A (en) | A kind of unzoned lens fringe region deblurring method based on symmetrical edge filling | |
| CN113658061A (en) | Hybrid constraint condition-oriented blurred image restoration method and terminal | |
| CN206864472U (en) | Imaging sensor |
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 | ||
| CB02 | Change of applicant information |
Address after: 9-2 Tangming Avenue, Guangming New District, Shenzhen City, Guangdong Province Applicant after: TCL China Star Optoelectronics Technology Co.,Ltd. Address before: 9-2 Tangming Avenue, Guangming New District, Shenzhen City, Guangdong Province Applicant before: Shenzhen China Star Optoelectronics Technology Co.,Ltd. |
|
| CB02 | Change of applicant information | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |