CN108765308A - A kind of image de-noising method based on convolution mask - Google Patents
A kind of image de-noising method based on convolution mask Download PDFInfo
- Publication number
- CN108765308A CN108765308A CN201810377220.4A CN201810377220A CN108765308A CN 108765308 A CN108765308 A CN 108765308A CN 201810377220 A CN201810377220 A CN 201810377220A CN 108765308 A CN108765308 A CN 108765308A
- Authority
- CN
- China
- Prior art keywords
- image
- filtering
- noising method
- convolution
- pretreatment
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 36
- 238000001914 filtration Methods 0.000 claims abstract description 50
- 238000004364 calculation method Methods 0.000 claims description 10
- 238000010606 normalization Methods 0.000 claims description 8
- 239000000284 extract Substances 0.000 abstract description 15
- 238000010586 diagram Methods 0.000 description 5
- 238000013507 mapping Methods 0.000 description 4
- 238000000605 extraction Methods 0.000 description 3
- 230000000717 retained effect Effects 0.000 description 3
- 230000002146 bilateral effect Effects 0.000 description 2
- 230000005540 biological transmission Effects 0.000 description 2
- 230000009466 transformation Effects 0.000 description 2
- 230000003044 adaptive effect Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 238000013139 quantization Methods 0.000 description 1
- 230000003313 weakening 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
- G06T5/70—Denoising; Smoothing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/13—Edge detection
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Image Processing (AREA)
Abstract
The present invention relates to a kind of image de-noising methods based on convolution mask, including：Original image is pre-processed, pretreatment image is obtained；Convolution denoising is carried out to the pretreatment image, obtains output image；The output image is modified, denoising image is obtained.The present embodiment extracts the characteristic information of image in such a way that convolution mask carries out convolution to image, during filtering and noise reduction, for the more complete of marginal information reservation, and completed by convolution operation, complicated formula does not calculate, and overall simple is convenient for hardware realization.
Description
Technical field
The invention belongs to image processing fields, and in particular to a kind of image de-noising method based on convolution mask.
Background technology
Image denoising refers to the process of reducing noise in image.In the transmission and acquisition process of image, often due to work
Make the influence of the factors such as environmental condition so that image is by noise jamming, so that the partial information of image is destroyed, Ren Leicong
The information extracted in image is also restrained.These noises may generate in the transmission, it is also possible to be generated in the processing such as quantization.
For piece image in practical applications there may be various noises, noise is the major reason of image interference.
The use of wide denoising method is at present two-sided filter denoising.Two-sided filter uses the Gauss of Euclidean distance
The gaussian kernel function product of kernel function and value differences is as filtering weighting filtering and noise reduction.This method is for big ladder in image
Marginal information preferable reservation again is spent, therefore can effectively filter out the noise in image, while retaining the marginal information of image.
However, using two-sided filter denoising, although image border can be retained, the joining place up and down at edge is not
It is bonded very much real image edge gradient, bad for the weakening effect of isolated noise point and calculating process using bilateral filtering is multiple
It is miscellaneous, it is unfavorable for hardware realization.
Invention content
In order to solve the above-mentioned problems in the prior art, the present invention provides a kind of images based on convolution mask to go
Method for de-noising.The technical problem to be solved in the present invention is achieved through the following technical solutions：
An embodiment of the present invention provides a kind of image de-noising methods based on convolution mask, including：
S1, original image is pre-processed, obtains pretreatment image；
S2, convolution denoising is carried out to the pretreatment image, obtains output image；
S3, the output image is modified, obtains denoising image.
In one embodiment of the invention, step S1 includes：
S11, mirror-extended is carried out to the original image outmost turns data, obtains pretreatment image.
In one embodiment of the invention, step S2 includes：
S21, weight coefficient is calculated；
S22, pass through Filtering Template and the pretreatment image, acquisition filtering image；
S23, the output image is obtained according to the weight coefficient and the filtering image.
In one embodiment of the invention, step S21 includes：
S211, pass through default template and the pretreatment image, acquisition first convolved image；
S212, normalization is carried out to first convolved image, obtains the second convolved image；
S213, weighted value is calculated according to second convolved image；
S214, the weight coefficient is calculated according to the weighted value.
In one embodiment of the invention, the default template in step S211 include 3*3 templates, 5*5 templates or
7*7 templates.
In one embodiment of the invention, step S22 includes：
S221, convolution is carried out by the Filtering Template and the pretreatment image, obtains filtering image.
In one embodiment of the invention, the Filtering Template is formed according to Gaussian template.
In one embodiment of the invention, step S3 includes：
S31, the variance for calculating the pretreatment image；
S32, setting threshold value；
S33, pass through the variance and the threshold calculations correction factor；
S34, according to the correction factor and the output image, utilize correction formula to obtain the denoising image.
In one embodiment of the invention, the correction factor calculation formula in step S33 is：
Wherein, a is the correction factor, and var (i, j) is the variance, and T is the threshold value.
In one embodiment of the invention, the correction formula in step S34 is：
Wherein, I_out (i, j) is the denoising image, and I_out1 (i, j) is the output image, I_out2 (i, j)
For the image for the original image obtain after mean filter, a is correction factor, and var (i, j) is the variance, and T is
The threshold value.
Compared with prior art, beneficial effects of the present invention：
(1) present invention proposition is a kind of carrying out convolution by convolution mask based on the image de-noising method of convolution mask to image
Mode extract the characteristic information of image, the filtering and noise reduction during, marginal information is retained more complete；
(2) present invention proposition is a kind of is completed based on the image de-noising method of convolution mask by convolution operation, not multiple
Miscellaneous formula calculates, and overall simple is convenient for hardware realization.
Description of the drawings
Fig. 1 is a kind of flow diagram of the image de-noising method based on convolution mask provided in an embodiment of the present invention；
Fig. 2 is the flow diagram of another image de-noising method based on convolution mask provided in an embodiment of the present invention.
Specific implementation mode
Further detailed description is done to the present invention with reference to specific embodiment, but embodiments of the present invention are not limited to
This.
Embodiment one
Fig. 1 is referred to, Fig. 1 is a kind of flow of the image de-noising method based on convolution mask provided in an embodiment of the present invention
Schematic diagram.The image de-noising method based on convolution mask, including：
S1, by being pre-processed to original image, obtain pretreatment image；
S2, by the pretreatment image carry out convolution denoising, obtain output image；
S3, by being modified to the output image, obtain denoising image.
Wherein, for step S1, may include：
S11, mirror-extended is carried out to the original image outmost turns data, obtains pretreatment image.
Wherein, for step S2, may include：
S21, weight coefficient is calculated；
S22, pass through Filtering Template and the pretreatment image, acquisition filtering image；
S23, the output image is obtained according to the weight coefficient and the filtering image.
Wherein, for step S21, may include：
S211, first convolved image is obtained by default template and the pretreatment image；
S212, normalization is carried out to first convolved image, obtains the second convolved image；
S213, weighted value is calculated according to second convolved image；
S214, the weight coefficient is calculated according to the weighted value.
Preferably, the default template in step S211 includes 3*3 templates, 5*5 templates and 7*7 templates.
Wherein, for step S22, may include：
S221, convolution is carried out by the Filtering Template and the pretreatment image, obtains filtering image.
Preferably, the Filtering Template in step S22 is formed according to Gaussian template, other modes can also be used to give birth to
At, for example, bilateral filtering template or according to the difference value of each point pixel value and local center pixel value come adaptive generation.
Wherein, for step S3, may include：
S31, the variance for calculating the pretreatment image；
S32, setting threshold value；
S33, pass through the variance and the threshold calculations correction factor；
S34, according to the correction factor and the output image, utilize correction formula to obtain the denoising image.
Preferably, the correction factor calculation formula in step S33 is：
Wherein, a is the correction factor, and var (i, j) is the variance, and T is the threshold value.
Preferably, the correction formula in step S34 is：
Wherein, I_out (i, j) is the denoising image, and I_out1 (i, j) is the output image, I_out2 (i, j)
For the image for the original image obtain after mean filter, a is correction factor, and var (i, j) is the variance, and T is
The threshold value.
The present embodiment extracts the characteristic information of image in such a way that convolution mask carries out convolution to image, is gone in filtering
During making an uproar, for the more complete of marginal information reservation, and completed by convolution operation, not complicated formula meter
It calculates, overall simple is convenient for hardware realization.
Embodiment two
It please join Fig. 2, Fig. 2 is the flow of another image de-noising method based on convolution mask provided in an embodiment of the present invention
Schematic diagram.The present embodiment on the basis of the above embodiments, is further described in detail the image de-noising method,
In, which specifically includes following steps so that size is the default template of 3*3 as an example：
Step 1：Mirror-extended is carried out to original image outmost turns data；
Noise-containing original image is denoted as I, to the outmost turns data of original image (the first row, first row, last
Row, last row) mirror-extended is carried out, pretreatment image is obtained, I_input is denoted as.
Step 2：Extract the marginal information of pretreatment image；
Convolution operation is carried out by four default templates and I_input, extracts the marginal information of pretreatment image, four pre-
If template is as follows：
Wherein, cov1 is the default template for extracting horizontal direction marginal information, and cov2 is extraction vertical direction marginal information
Default template, cov3 be extract 45 degree direction marginal informations default template, cov4 be extract 135 degree of direction marginal informations
Default template.
Preferably, the first convolved image is obtained by default template, wherein the first convolved image be denoted as respectively I1, I2,
I3、I4。
Step 3：Normalization is carried out to the first convolved image；
Cov1 and cov2 has used three pairs of pixel information in step 2, and cov3 and cov4 have only used two pairs of pixels
Information needs normalization of progress to obtain the second convolved image.Wherein, for some pixel (i, j), pass through following public affairs
Formula carries out normalization：
I1'(i, j)=I1 (i, j)/3, I2'(i, j)=I2 (i, j)/3,
I3'(i, j)=I3 (i, j)/2, I4'(i, j)=I4 (i, j)/2,
Wherein, I1'(i, j), I2'(i, j), I3'(i, j) and, I4'(i, j) it is respectively corresponding second convolution image pixel point
Information, I1 (i, j), I2 (i, j), I3 (i, j), I4 (i, j) are respectively corresponding first convolution image pixel point information.
Step 4：Weighted value is calculated according to second convolved image；
According to step 3 as a result, compare I1'(i, j), I2'(i, j), I3'(i, j), I4'(i, j) four values,
If four values are all 0, weighted value w1(i, j)=w2(i, j)=w3(i, j)=w4(i, j)=0.25；
Otherwise, it is as follows to calculate weighted value formula：
Wherein, w1(i, j), w2(i, j), w3(i, j), w4(i, j) is respectively corresponding weighted value.
Step 5：The weight coefficient is calculated according to weighted value；
Nonlinear Mapping transformation is carried out to calculated weighted value in step 4, widens the gap between weighted value.Specifically
Nonlinear Mapping process calculation formula is as follows：
Wherein, w '1(i, j), w'2(i, j), w'3(i, j) and w'4(i, j) is respectively corresponding weight coefficient, and γ is adjustment
Parameter (γ >=1), as γ=1, weight coefficient is not stretched.
Step 6：Carry out convolution denoising；
Filtering Template is generated by Gaussian template, wherein Gaussian template is obtained by Gaussian function, and Gaussian template here
Parameter selection σ=1, then all directions it is as follows to correspond to Filtering Template：
Wherein, g1 is horizontal direction Filtering Template, and g2 is vertical direction Filtering Template, and g3 is 45 degree of trend pass filtering templates,
G4 is 135 degree of trend pass filtering templates.
Preferably, four corresponding filtering figures are obtained by Filtering Template g1, g2, g3, g4 convolution corresponding to I_input progress
As I1 ", I2 ", I3 ", I4 ".
Preferably, output image is obtained by filtering image and weight coefficient, wherein be directed to certain pixel (i, j), meter
It is as follows to calculate formula：
I_out1 (i, j)=w1' (i, j) × I1 " (i, j)+w'2(i, j) × I2 " (i, j)+w'3(i, j) × I3 " (i, j)+
w'4(i, j) × I4 " (i, j)
Wherein, I_out1 is output image.
Step 7：Output image is modified, denoising image is obtained；
Calculate the variance var, given threshold T, according to variance and threshold value of the local 3*3 windows of pretreatment image I_input
Relationship to output image be modified, correction formula is as follows：
Wherein, I_out (i, j) is the denoising image, and I_out1 (i, j) is the output image, I_out2 (i, j)
For the image for the original image obtain after mean filter, a is correction factor, and var (i, j) is the variance, and T is
The threshold value.
Preferably, mean filter refers to all pixels in the regional area centered on some pixel (i, j), is constituted
Then one Filtering Template removes pixel (i, j) itself, seeks the average value of all pixels in regional area, then uses average value
To replace the pixel value of (i, j).
Preferably, the filter of the size selection 7*7 of the corresponding mean filters of I_out2 (i, j).And correction factor a
Calculation formula is：
The image de-noising method that the present embodiment proposes makes to extract image in such a way that different convolution masks are combined
Marginal information is more reasonable, more accurately, and by the way that Gaussian template to be combined with coefficient distribution, ensure that the weight of noise spot
Contribution is smaller, and the weight contribution of effective information point is larger；Finally in the local variance by threshold value and calculating image, according to setting
Threshold value further to carry out denoising to smooth region, denoising effect is more preferable.
Embodiment three
Continuing with referring to Fig. 2, Fig. 2 is another image de-noising method based on convolution mask provided in an embodiment of the present invention
Flow diagram.The present embodiment on the basis of the above embodiments, further to the image de-noising method retouch in detail
It states, wherein the image de-noising method specifically includes following steps so that size is the default template of 5*5 as an example：
Step 1：Mirror-extended is carried out to original image outmost turns data；
Noise-containing original image is denoted as I, to the outmost turns data of original image (the first row, first row, last
Row, last row) mirror-extended is carried out, pretreatment image is obtained, I_input is denoted as.
Step 2：Extract the marginal information of pretreatment image；
Convolution operation is carried out by eight default templates and I_input, extracts the marginal information of pretreatment image, eight pre-
If template is as follows：
Wherein, cov1 is the default template for extracting horizontal direction marginal information, and cov2 is extraction vertical direction marginal information
Default template, cov3 be extract 45 degree direction marginal informations default template, cov4 be extract 135 degree of direction marginal informations
Default template wherein, cov5 be extract 22.5 degree direction marginal informations default template, cov6 be extract 67.5 degree of directions side
The default template of edge information, cov7 are the default template for extracting 112.5 degree of direction marginal informations, and cov8 is 157.5 degree of sides of extraction
To the default template of marginal information, obtained convolved image is denoted as I1, I2, I3, I4, I5, I6, I7, I8 respectively.
Step 3：Normalization is carried out to the first convolved image；
In step 2 kind, since cov1 and cov2 has used five pairs of pixel information, and cov3 and cov4 are only used
Four pairs of pixel information, cov5, cov6, cov7, cov8 have only used three pairs of pixel information, therefore are directed to some pixel
(i, j), carries out normalization, and formula is as follows：
I1'(i, j)=I1 (i, j)/5, I2'(i, j)=I2 (i, j)/5,
I3'(i, j)=I3 (i, j)/4, I4'(i, j)=I4 (i, j)/4,
I5'(i, j)=I5 (i, j)/3, I6'(i, j)=I6 (i, j)/3,
I7'(i, j)=I7 (i, j)/3, I8'(i, j)=I8 (i, j)/3,
Step 4：Weighted value is calculated according to second convolved image；
Compare I1'(i, j for wherein certain pixel (i, j) according to the result of calculation of step 3), I2'(i, j), I3'
(i, j), I4'(i, j), I5'(i, j), I6'(i, j), I7'(i, j), I8'(i, j) this eight value, if this eight value all be 0,
Have：
w1(i, j)=w2(i, j)=w3(i, j)=w4(i, j)=w5(i, j)=w6(i, j)=w7(i, j)=w8(i, j)=
0.125
Otherwise, it is as follows to calculate weight equation：
Wherein, w1(i, j), w2(i, j), w3(i, j), w4(i, j) w5(i, j), w6(i, j), w7(i, j), w8(i, j), respectively
For corresponding weighted value.
Step 5：The weight coefficient is calculated according to weighted value；
Nonlinear Mapping transformation is carried out to calculated weighted value in step 4, widens the gap between weighted value.Specifically
Nonlinear Mapping process calculation formula is as follows：
Wherein, w '1(i, j), w'2(i, j), w'3(i, j) and w'4(i, j) is respectively corresponding weight coefficient, and γ is adjustment
Parameter (γ >=1), as γ=1, weight coefficient is not stretched.
Step 6：Carry out convolution denoising；
Filtering Template is generated by Gaussian template, wherein Gaussian template is obtained by Gaussian function, and Gaussian template here
Parameter selection σ=1, then all directions it is as follows to correspond to Filtering Template：
Wherein, g1 is horizontal direction Filtering Template, and g2 is vertical direction Filtering Template, and g3 is 45 degree of trend pass filtering templates,
G4 is 135 degree of trend pass filtering templates.G5 is 22.5 degree of Filtering Templates, and g6 is 67.5 degree of trend pass filtering templates, and g7 is 112.5 degree
Trend pass filtering template, g8 are 157.5 degree of trend pass filtering templates.
Preferably, pass through Filtering Template g1、g2、g3、g4、g5、g6、g7、g8Convolution corresponding to I_input progress obtains eight
Corresponding filtering image I1 ", I2 ", I3 ", I4 ", I5 ", I6 ", I7 ", I8 ".
Preferably, it crosses filtering image and weight coefficient obtains output image, wherein be directed to certain pixel (i, j), calculate
Formula is as follows：
I_out1 (i, j)=w1' (i, j) × I1 " (i, j)+w'2(i, j) × I2 " (i, j)+w'3(i, j) × I3 " (i, j)+
w'4(i, j) × I4 " (i, j)
Wherein, I_out1 is output image.
Step 7：Output image is modified, denoising image is obtained；
Calculate the variance var, given threshold T, according to variance and threshold value of 5 × 5 windows of part of pretreatment image I_input
Relationship to output image be modified, correction formula is as follows：
Wherein, I_out (i, j) is the denoising image, and I_out1 (i, j) is the output image, I_out2 (i, j)
For the image for the original image obtain after mean filter, a is correction factor, and var (i, j) is the variance, and T is
The threshold value.
Preferably, mean filter refers to all pixels in the regional area centered on some pixel (i, j), is constituted
Then one Filtering Template removes pixel (i, j) itself, seeks the average value of all pixels in regional area, then uses average value
To replace the pixel value of (i, j).
Preferably, the filter of the size selection 7*7 of the corresponding mean filters of I_out2 (i, j).And correction factor a
Calculation formula is：
The image de-noising method that the present embodiment proposes extracts image in such a way that convolution mask carries out convolution to image
Characteristic information, different convolution masks extracts different characteristic information, and a series of convolution masks combine, and use can be so that carry
The feature taken is more perfect, more comprehensively.Therefore during filtering and noise reduction, marginal information is retained more complete.
The above content is a further detailed description of the present invention in conjunction with specific preferred embodiments, and it cannot be said that
The specific implementation of the present invention is confined to these explanations.For those of ordinary skill in the art to which the present invention belongs, exist
Under the premise of not departing from present inventive concept, a number of simple deductions or replacements can also be made, all shall be regarded as belonging to the present invention's
Protection domain.
Claims (10)
1. a kind of image de-noising method based on convolution mask, which is characterized in that including：
S1, original image is pre-processed, obtains pretreatment image；
S2, convolution denoising is carried out to the pretreatment image, obtains output image；
S3, the output image is modified, obtains denoising image.
2. image de-noising method according to claim 1, which is characterized in that step S1 includes：
S11, mirror-extended is carried out to the original image outmost turns data, obtains pretreatment image.
3. image de-noising method according to claim 1, which is characterized in that step S2 includes：
S21, weight coefficient is calculated；
S22, pass through Filtering Template and the pretreatment image, acquisition filtering image；
S23, the output image is obtained according to the weight coefficient and the filtering image.
4. image de-noising method according to claim 3, which is characterized in that step S21 includes：
S211, pass through default template and the pretreatment image, acquisition first convolved image；
S212, normalization is carried out to first convolved image, obtains the second convolved image；
S213, weighted value is calculated according to second convolved image；
S214, the weight coefficient is calculated according to the weighted value.
5. image de-noising method according to claim 4, which is characterized in that the default template in step S211 includes
3*3 templates, 5*5 templates or 7*7 templates.
6. image de-noising method according to claim 3, which is characterized in that step S22 includes：
S221, convolution is carried out by the Filtering Template and the pretreatment image, obtains filtering image.
7. image de-noising method according to claim 3, which is characterized in that form the filtering mould according to Gaussian template
Plate.
8. image de-noising method according to claim 1, which is characterized in that step S3 includes：
S31, the variance for calculating the pretreatment image；
S32, setting threshold value；
S33, pass through the variance and the threshold calculations correction factor；
S34, according to the correction factor and the output image, utilize correction formula to obtain the denoising image.
9. image de-noising method according to claim 8, which is characterized in that the correction factor in step S33 calculates
Formula is：
Wherein, a is the correction factor, and var (i, j) is the variance, and T is the threshold value.
10. image de-noising method according to claim 9, which is characterized in that the correction formula in step S34 is：
Wherein, I_out (i, j) is the denoising image, and I_out1 (i, j) is the output image, and I_out2 (i, j) is to institute
It states original image and carries out the image obtained after mean filter, a is correction factor, and var (i, j) is the variance, and T is the threshold
Value.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810377220.4A CN108765308B (en) | 2018-04-25 | 2018-04-25 | Image denoising method based on convolution template |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810377220.4A CN108765308B (en) | 2018-04-25 | 2018-04-25 | Image denoising method based on convolution template |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108765308A true CN108765308A (en) | 2018-11-06 |
| CN108765308B CN108765308B (en) | 2022-02-18 |
Family
ID=64011891
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810377220.4A Active CN108765308B (en) | 2018-04-25 | 2018-04-25 | Image denoising method based on convolution template |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108765308B (en) |
Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1455368A (en) * | 2003-04-09 | 2003-11-12 | 重庆大学 | Digital image processing method for restoring reduced quality of image space shift |
| CN102222326A (en) * | 2011-06-28 | 2011-10-19 | 青岛海信信芯科技有限公司 | Method and device for deblurring images based on single low resolution |
| CN103067647A (en) * | 2012-12-25 | 2013-04-24 | 四川九洲电器集团有限责任公司 | Field programmable gata array (FPGA) based video de-noising method |
| CN103077508A (en) * | 2013-01-25 | 2013-05-01 | 西安电子科技大学 | Transform domain non local and minimum mean square error-based SAR (Synthetic Aperture Radar) image denoising method |
| CN104200442A (en) * | 2014-09-19 | 2014-12-10 | 西安电子科技大学 | Improved canny edge detection based non-local means MRI (magnetic resonance image) denoising method |
| CN105913393A (en) * | 2016-04-08 | 2016-08-31 | 暨南大学 | Self-adaptive wavelet threshold image de-noising algorithm and device |
| 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 |
| CN107392314A (en) * | 2017-06-30 | 2017-11-24 | 天津大学 | A kind of deep layer convolutional neural networks method that connection is abandoned based on certainty |
| CN107633486A (en) * | 2017-08-14 | 2018-01-26 | 成都大学 | Structure Magnetic Resonance Image Denoising based on three-dimensional full convolutional neural networks |
| WO2018018470A1 (en) * | 2016-07-27 | 2018-02-01 | 华为技术有限公司 | Method, apparatus and device for eliminating image noise and convolutional neural network |
-
2018
- 2018-04-25 CN CN201810377220.4A patent/CN108765308B/en active Active
Patent Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1455368A (en) * | 2003-04-09 | 2003-11-12 | 重庆大学 | Digital image processing method for restoring reduced quality of image space shift |
| CN102222326A (en) * | 2011-06-28 | 2011-10-19 | 青岛海信信芯科技有限公司 | Method and device for deblurring images based on single low resolution |
| CN103067647A (en) * | 2012-12-25 | 2013-04-24 | 四川九洲电器集团有限责任公司 | Field programmable gata array (FPGA) based video de-noising method |
| CN103077508A (en) * | 2013-01-25 | 2013-05-01 | 西安电子科技大学 | Transform domain non local and minimum mean square error-based SAR (Synthetic Aperture Radar) image denoising method |
| CN104200442A (en) * | 2014-09-19 | 2014-12-10 | 西安电子科技大学 | Improved canny edge detection based non-local means MRI (magnetic resonance image) denoising method |
| CN105913393A (en) * | 2016-04-08 | 2016-08-31 | 暨南大学 | Self-adaptive wavelet threshold image de-noising algorithm and device |
| 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 |
| WO2018018470A1 (en) * | 2016-07-27 | 2018-02-01 | 华为技术有限公司 | Method, apparatus and device for eliminating image noise and convolutional neural network |
| CN107392314A (en) * | 2017-06-30 | 2017-11-24 | 天津大学 | A kind of deep layer convolutional neural networks method that connection is abandoned based on certainty |
| CN107633486A (en) * | 2017-08-14 | 2018-01-26 | 成都大学 | Structure Magnetic Resonance Image Denoising based on three-dimensional full convolutional neural networks |
Non-Patent Citations (1)
| Title |
|---|
| 李传朋 等: "基于深度卷积神经网络的图像去噪研究", 《计算机工程》 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108765308B (en) | 2022-02-18 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Zhou et al. | Retinex-based laplacian pyramid method for image defogging | |
| Singh et al. | Evolving fusion-based visibility restoration model for hazy remote sensing images using dynamic differential evolution | |
| CN103236040B (en) | A kind of color enhancement method and device | |
| CN114140353A (en) | Swin-Transformer image denoising method and system based on channel attention | |
| CN109389560B (en) | Adaptive weighted filtering image noise reduction method and device and image processing equipment | |
| CN114066747B (en) | Low-illumination image enhancement method based on illumination and reflection complementarity | |
| CN110390646B (en) | Detail-preserving image denoising method | |
| Liu et al. | Learning hadamard-product-propagation for image dehazing and beyond | |
| CN104182947A (en) | Low-illumination image enhancement method and system | |
| CN109064423A (en) | It is a kind of based on unsymmetrical circulation generate confrontation loss intelligence repair drawing method | |
| CN107492075A (en) | A kind of method of individual LDR image exposure correction based on details enhancing | |
| Zhang et al. | Exploiting image local and nonlocal consistency for mixed Gaussian-impulse noise removal | |
| CN102298774A (en) | Non-local mean denoising method based on joint similarity | |
| CN104616259B (en) | A kind of adaptive non-local mean image de-noising method of noise intensity | |
| CN117252773A (en) | Image enhancement method and system based on self-adaptive color correction and guided filtering | |
| CN114202460B (en) | Super-resolution high-definition reconstruction method, system and equipment for different damage images | |
| Lei et al. | A novel intelligent underwater image enhancement method via color correction and contrast stretching✰ | |
| Yang et al. | Joint image dehazing and super-resolution: Closed shared source residual attention fusion network | |
| Chen et al. | CERL: A unified optimization framework for light enhancement with realistic noise | |
| CN111652809A (en) | Infrared image noise suppression method for enhancing details | |
| CN107590781B (en) | Self-adaptive weighted TGV image deblurring method based on original dual algorithm | |
| CN104966271B (en) | Image de-noising method based on biological vision receptive field mechanism | |
| CN108765308A (en) | A kind of image de-noising method based on convolution mask | |
| Tun et al. | Joint Training of Noisy Image Patch and Impulse Response of Low-Pass Filter in CNN for Image Denoising | |
| Zhu et al. | LLISP: Low-light image signal processing net via two-stage network |
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 |