CN108765332A - A kind of non-local mean denoising method of ellipse search window and parameter adaptive - Google Patents
A kind of non-local mean denoising method of ellipse search window and parameter adaptive Download PDFInfo
- Publication number
- CN108765332A CN108765332A CN201810502466.XA CN201810502466A CN108765332A CN 108765332 A CN108765332 A CN 108765332A CN 201810502466 A CN201810502466 A CN 201810502466A CN 108765332 A CN108765332 A CN 108765332A
- Authority
- CN
- China
- Prior art keywords
- image
- search window
- oval
- value
- pixel
- 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 29
- 230000003044 adaptive effect Effects 0.000 title claims abstract description 25
- 238000009499 grossing Methods 0.000 claims abstract description 14
- 238000004458 analytical method Methods 0.000 claims abstract description 6
- 238000009825 accumulation Methods 0.000 claims description 6
- 238000000354 decomposition reaction Methods 0.000 claims description 6
- 239000011159 matrix material Substances 0.000 claims description 4
- 238000013461 design Methods 0.000 claims description 3
- 230000003321 amplification Effects 0.000 claims description 2
- 230000001186 cumulative effect Effects 0.000 claims description 2
- 238000010606 normalization Methods 0.000 claims description 2
- 238000003199 nucleic acid amplification method Methods 0.000 claims description 2
- 230000014509 gene expression Effects 0.000 claims 5
- 230000000694 effects Effects 0.000 abstract description 13
- 238000003672 processing method Methods 0.000 abstract description 3
- 238000002474 experimental method Methods 0.000 abstract description 2
- 238000012545 processing Methods 0.000 abstract description 2
- 238000012360 testing method Methods 0.000 description 9
- 241000721047 Danaus plexippus Species 0.000 description 8
- 230000006978 adaptation Effects 0.000 description 2
- 230000007547 defect Effects 0.000 description 1
- 230000007812 deficiency Effects 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 230000001737 promoting effect Effects 0.000 description 1
- 230000000717 retained effect Effects 0.000 description 1
- 230000003595 spectral effect Effects 0.000 description 1
Classifications
-
- G06T5/70—
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/40—Image enhancement or restoration by the use of histogram techniques
Abstract
The present invention relates to a kind of image denoising processing methods based on non-local mean frame, it is related to image processing techniques, using the oval search window consistent with image local area structure, according to the partial structurtes of image, the progress of smoothing parameter numerical value in the size and Denoising Algorithm of oval search window is adaptively adjusted, the gray value preferably to treat denoising pixel is estimated.Denoising effect of the present invention under different noise circumstances has relatively good robustness.The analysis that the present invention passes through histogram information and image array information to image local area, realize that the tile size based on non-local mean algorithm is adaptive, smoothing parameter numerical value is adaptive and search window form adaptive, to effectively carry out noise suppressed to image detail part and retain the texture information of detail section as much as possible, realize the improvement to traditional non-local mean algorithm, it is apparent that experiment effect proves that innovatory algorithm is promoted in denoising effect and texture part.
Description
Technology neighborhood
The present invention relates to image processing techniques neighborhoods, more particularly to a kind of image denoising processing method, can be used for certainly
Gaussian noise present in right image is handled.
Background technology
In the noise suppressed of digital picture and the direction of denoising, non-local mean Denoising Algorithm (NLM) utilizes Gauss
The characteristics of there are other similar image blocks in property and image that white noise mean value is zero, by searching for it in local neighborhood
His similar pixel point carries out weighted average to estimate the actual value of calculating target pixel points.The algorithm has preferable denoising effect
There are many improved methods to be proposed for promoting the property of NLM since proposition with the ability for preserving image detail part
Can, it is improved because being known as example:Computational efficiency, the shape of search window, the size adaptation of image block are weighting function, smooth
Coefficient it is adaptive, from weighted value and the method etc. that passes through prior information.
There are some defects in traditional non-local mean Denoising Algorithm (NLM).It is as follows:(1) although NLM has centainly
Denoising effect, but with the increase of noise grade, the reserve capability of image detail part is degenerated seriously, and detail section includes
Many important image informations;(2) tradition NLM algorithms put on an equal footing each pixel in image, cannot be effectively
Pixel with similar features in image is distinguished and is analyzed.
Invention content
It is an object of the invention to improve the deficiency in traditional NLM algorithms, it is proposed that a kind of improved image denoising processing
Method realizes a kind of image block based on NLM algorithms by the analysis of histogram information and matrix information to image local
Size adaptation, smoothing parameter numerical value is adaptive and search window form adaptive algorithm.
To achieve the goals above, the present invention provides following technical scheme, particular content is as follows:
(1) one pair of input waits for the noise image I of denoising;
(2) noise image is stepped through, the rectangular local neighborhood Ω centered on pixel i is obtainedi;
(3) gradient image G is calculatedi, the structure tensor C of each pixel i is then obtained according to gradient imagei;
(4) to structure tensor CiIt carries out singular value decomposition and obtains feature vector Fi;
(5) rectangular local neighborhood Ω is calculatediHistogram second moment
(6) by local neighborhood ΩiHistogram second momentWith the characteristic value S of gradient imagei, it is calculated pixel i's
Classification indicators value Ri, obtain the classification indicators value R of image I all pixelsiThe classification chart I formedR;
(7) classification chart I is calculatedRFrequency accumulation histogram T, each pixel RiCompare T in frequency be 30%, 70%,
Classification indicators at 90% are worth to the classification belonging to pixel i, and wherein frequency refers to classification indicators value;
(8) based on designed nonlinear function, the tile size of each pixel i in non-local mean algorithm is obtained
Parameter psiAnd smoothing parameter hi;
(9) by the feature vector F of structure tensori, the direction of rotation of oval search window is calculated, by the spy of gradient image
Value indicative Si, the long axis and short axle size of oval search window is calculated, so that it is determined that the parametric equation E of oval search windowi(z);
(10) rectangular local neighborhood Ω is traversediInterior all pixels point calculates between oval search window internal image block
Weighted value;
(11) image block after denoising is calculated
(12) (2)~(11) are repeated, until all pixels point traversal is completed in noise image;
(13) method for using arithmetic mean for the estimation image block of overlapping, is finally calculated the image after denoising.
Compared with prior art, the present invention has the following advantages:
The adaptive, smoothing parameter by the analysis realization tile size to image local histogram information and matrix information
Numerical value is adaptive and search window form adaptive, to effectively carry out noise suppressed to image detail part, simultaneously
The pattern and texture of detail section can be retained as much as possible again.
Below by drawings and examples, technical scheme of the present invention will be described in further detail.
Description of the drawings
Fig. 1 is the flow chart of the embodiment of the present invention;
The example pixel point that Fig. 2 chooses in the Monarch test charts for the present invention;
The histogram of Square Neighborhood where the example pixel point that Fig. 3 chooses in the Monarch test charts for the present invention;
Fig. 4 is the nonlinear function curve graph of the embodiment of the present invention;
Fig. 5 be the embodiment of the present invention Lena test charts in different zones classification schematic diagram;Wherein, (a) is that nothing is made an uproar
The Lena test images of sound are (b) the Lena images under the noise grades of σ=20, four are divided the image into according to the difference of color
A region:Flat site (black), strong noise texture (light gray), intermediate region (white) and low noise texture region are (dark-grey
Color);
Fig. 6 be the embodiment of the present invention Lena test charts in multiple points oval search window region;
Fig. 7 is eight standard testing images (512 × 512) used in the embodiment of the present invention;
Fig. 8 is Monarch standard testings image used in the embodiment of the present invention and uses distinct methods denoising effect
Figure;Wherein, (a) is muting Monarch images;(b) it is the Monarch images under the noise grades of σ=20;(c) it is to use
Result figure after traditional NLM denoisings;(d) it is to use the result figure after ANLM denoisings;(e) it is the result after the method for the present invention denoising
Figure.
Specific implementation mode
The embodiment of the present invention is illustrated below in conjunction with attached drawing, it should be understood that embodiment described herein is only used
In the description and interpretation present invention, it is not intended to limit the present invention.
Embodiment
The non-local mean denoising method of a kind of oval search window and parameter adaptive provided in this embodiment is specific to wrap
Following steps are included, referring to Fig. 1:
(1) the Monarch noise image I of a pair 512 × 512 are inputted, noise grade is σ=20.
(2) the pixel i in noise image I is stepped through, the pixel at image (357,354) is had chosen in this example
Point conduct
Example illustrates, and sees Fig. 2, obtains the local neighborhood Ω centered on the pixeli, Size of Neighborhood is 11 × 11.
(3) rectangular local neighborhood Ω is calculated by following formula 1iAlong gradient image G both horizontally and verticallyi:
Then, to gradient image GiIt carries out singular value decomposition and obtains characteristic value Si=(Si,1,Si,2) and feature vector Vi=
(Vi,1,Vi,2), i.e.,
Structure tensor C is calculated by following formula 2i:
Wherein,Related with the size of analysis window for amplification coefficient, M=121 is that rectangular part is adjacent
Domain Ωi
The total number of middle pixel, σiIt is characterized value Si,1And Si,2Between ratio σi=Si,1/Si,2=2.1694.
(4) to structure tensor CiIt carries out singular value decomposition and feature vector F is calculatedi,1,Fi,2;
Fi=[Fi,1,Fi,2]
(5) rectangular local neighborhood Ω is calculated by formula 3iHistogram second momentHistogram is shown in Fig. 3:
Wherein, M=11 × 11.
(6) by local neighborhood ΩiHistogram second momentWith the characteristic value S of structure tensori, it is calculated pixel i's
Classification indicators value Ri.The classification indicators value R of pixel i is calculated according to following formula 4i, it is calculated by classification indicators value structure
At classification chart IR:
(7) by following formula 5 according to classification indicators value RiIn classification chart IRFrequency accumulation histogram T in where frequency
Threshold range is sorted out, and the classification belonging to pixel i is obtained.
Wherein classification chart IR30%, 70% and 90% places that arrange from small to large frequency accumulative histogram T it is corresponding
Classification indicators value T0.3=0.0162, T0.7=0.0812 and T0.9=0.3735 is divided into low noise texture region c1, intermediate region c2、
Strong noise texture region c3With flat site c4Four classifications, due to RiBetween T0.7And T0.9Between, therefore pixel is assigned into centre
Region.
(8) different size of image block ps is defined to all pixels returned in category regions in imagei, flat site
c4Tile size be 11 × 11, intermediate region c2Tile size be 9 × 9, strong noise texture region c3Image block it is big
It is small be 5 × 5, low noise texture region c1Tile size be 7 × 7.
(9) according to nonlinear function f1, f2And f3, referring to Fig. 2, the smoothing factor h of pixel i is calculatedi。
With psiThe size for indicating image block at pixel i, h is calculated by following formula 6i, to indicate that pixel i is making an uproar
Smoothing factor value in acoustic image I:
Wherein,For the design factor of chi square distribution,It indicates
Degree of freedom is psiAnd probability value is the chi-square value corresponding to 0.99.
(10) to structure tensor CiIt carries out spectral factorization and obtains feature vector Fi,1And Fi,2, it is inclined that oval search window is calculated
Angle, θ from Y-axis positive directioniIf θiFor just, then to rotating clockwise, if θiFor negative, then rotate counterclockwise.Under
It states formula 7 and elliptical direction of rotation θ is calculatedi:
By characteristic value Si,1And Si,2, the long axis and short axle of oval search window are calculated according to following formula 8.
By obtaining the long axis and short axle of oval search window:
(11) local direction neighborhood Ω is calculatediIn the distance between other pixels and oval center pixel point, calculate
To the equation E of oval search windowi(z)。
First rectangular local neighborhood Ω is calculated by following formula 9iIn coordinate between other pixels and central point i away from
From:
Wherein, Δ x and Δ y indicates other coordinate differences of point with elliptical center point in oval search window.
Then the elliptic equation E of oval search window is obtained by following formula 10i(z):
(12) rectangular local neighborhood Ω is traversediInterior all pixel j in addition to i, if traversal point is in oval search window
The inside of mouth equation (will the postrotational (x of traversal pointr,yr) coordinate is brought elliptic equation into and judged, if Ei(z)≤1 determination
For inside ellipse or on ellipse), that is, meet Ei(z)≤1 it, then calculates the image block centered on the point and the oval center of circle is
Weighted value W between the image block at centerijIf traversal point is not in the inside of oval search window equation, i.e. Ei(z) 1 > then weighs
Duplicate step is until ΩiInterior all pixels point has traversed.
(13) in oval search window ΦiInternal pixel, by following formula 11 be calculated traversal point with it is ellipse
Weighted value W between circle centre pointij。
The image block P in oval search window centered on other pixels j is calculated by following formula 11jJustify with ellipse
Image block P centered on heart point iiBetween weighted value Wij:
Wherein, ZiFor normalization factor;hiFor smoothing factor, it is calculated by formula (5);ΦiIt is ellipse centered on i
Circle search window;Indicate the weighted euclidean distance between two image blocks.
(14) image block centered on the oval centre point i after denoising is calculated by following formula 12
(15) step (2)~(14), the image block after all pixels point denoising is calculated are repeated;
(16) pixel value of lap is cumulative is averaged, and finally obtains the Monarch images after denoising.
Fig. 5 is 8 test images, in the case where standard deviation is 20,25,30 and 35 noise image (512 × 512), is made respectively
Denoising is carried out with NLM, ANLM, the method for the present invention denoising method, test result is as shown in table 1.
Table 1:Denoising result contrast table
By table 1, as it can be seen that the method for the present invention is from objective image evaluation index PSNR (Y-PSNR), (structure is similar with SSIM
Degree) numerical value from the point of view of have better denoising effect.Denoising from Fig. 8 comparison distinct methods to noise image Monarch images
From the point of view of in effect, the pistil details profile after NLM algorithm process fogs, and ANLM is compared with NLM to be had in details reduction
Subtle promotion, and method using the present invention can see pistil part in image better reduction effect.Therefore, from
In subjective vision angle and objective indicator angle, the method for the present invention has better denoising effect and details reserve capability.
Referring to Fig. 4, the elliptic region of multiple points in figure, it can be seen that the form of search window is in flat site because being characterized
The small otherness of value is approximately a circle, and in the detail section of image, strong boundary window shape is the same picture of elongated strip
The structure feature of image matches where vegetarian refreshments.Using the oval search window consistent with image local area structure, according to figure
The partial structurtes of picture adaptively adjust the progress of smoothing parameter numerical value in the size and Denoising Algorithm of oval search window
Whole, the gray value preferably to treat denoising pixel is estimated.
As it can be seen that analysis of the present invention by histogram information and matrix information to image local, is realized based on non local
The tile size of mean algorithm is adaptive, smoothing parameter numerical value is adaptive and search window form adaptive, to effectively
Ground is carrying out noise suppressed to image detail part, while can retain the pattern and texture of detail section as much as possible again, real
Now to the improvement of traditional non-local mean algorithm;It is bright that experiment effect proves that innovatory algorithm is promoted on denoising effect and texture part
It is aobvious, there is significantly more efficient denoising effect, while having better reduction to the detail textures part of noise image;Using pixel
The corresponding tile size of point adaptively and the self-adapting regulation method of smoothing parameter numerical value, the image block after calculating estimation
Using oval window identical with image local area structure direction when pixel value, there is relatively good Shandong under different noise circumstances
Stick.
It should be appreciated that the above embodiment of the present invention and example, be to be not intended to limit this hair for description and interpretation purpose
Bright range.The scope of the present invention is defined by claim, rather than by above-described embodiment and example definition.
Claims (10)
1. a kind of non-local mean denoising method of ellipse search window and parameter adaptive, including the following contents:
(1) one pair of input waits for the noise image I of denoising;
(2) noise image is stepped through, the rectangular local neighborhood Ω centered on pixel i is obtainedi;
(3) rectangular local neighborhood Ω is calculatediGradient image Gi, the structure of each pixel i in field is obtained according to gradient image
Tensor Ci;
(4) to structure tensor CiIt carries out singular value decomposition and obtains the feature vector F of structure tensori;
(5) rectangular local neighborhood Ω is calculatediHistogram second moment
(6) by rectangular local neighborhood ΩiHistogram second momentWith the characteristic value D of structure tensori, it is calculated pixel i's
Classification indicators value Ri, obtain the classification indicators value R of image I all pixelsiThe classification chart I formedR;
(7) by IRFrequency accumulation histogram T is calculated, according to the frequency threshold value of classification, obtains the classification belonging to pixel i;
(8) nonlinear function based on design obtains the tile size parameter ps of each pixel i in non-local mean algorithmi
And smoothing parameter hi;
(9) by the feature vector F of structure tensori, the direction of rotation of oval search window is calculated;By gradient image GiFeature
Value Si, the long axis and short axle size of oval search window is calculated, so that it is determined that the parametric equation of oval search window;
(10) rectangular regional area Ω is traversediInterior all pixels point calculates the weight between oval search window internal image block
Value;
(11) image block after denoising is calculatedAnd all pixels point traversal is completed just to jump to next program in noise image
(12);If also having other pixels not traverse in noise image, program (2) is jumped to;
(12) estimated value of denoising image block is calculated, for the method that lap uses arithmetic mean, finally calculates
Image after to denoising.
2. a kind of non-local mean denoising method of oval search window and parameter adaptive according to claim 1, special
Sign is, in the step (3), calculates the rectangular local neighborhood Ω of noise image by following formula 1 firstiGradient image Gi:
Then to GiIt carries out singular value decomposition and obtains the characteristic value S of gradient imagei=(Si,1,SI, 2) and feature vector Vi=(Vi,1,
VI, 2), structure tensor is calculated by following formula 2:
Wherein,It is related with the size of analysis window for amplification coefficient;M is rectangular local neighborhood ΩiMiddle picture
The number of vegetarian refreshments, λ " are constant, σi=Si,1/SI, 2。
3. a kind of non-local mean denoising method of oval search window and parameter adaptive according to claim 1, special
Sign is, in the step (5), the second moment of accumulation histogramIt is calculated by following formula 3:
Wherein, yj is Square Neighborhood ΩiThe gray value of interior pixel j,For rectangular local neighborhood ΩiInterior all pixels point is averaged
Value, njIndicate pixel number identical with the gray value of pixel j in local neighborhood.
4. a kind of non-local mean denoising method of oval search window and parameter adaptive according to claim 1, special
Sign is, in the step (6), the classification indicators value R of pixel i is calculated by following formula 4i, RiIncluding local neighborhood
Matrix information and accumulation histogram information:
Wherein,The texture magnitude of local neighborhood is weighed, bigger expression pixel is located at texture area
Domain, smaller expression pixel are in non-grain region;For variance measures value, it is inversely proportional with noise variance;
RiThe bigger expression pixel i of value more levels off to the low-noise area of texture, and smaller expression pixel i more levels off to non-grain height
Noise.
5. a kind of non-local mean denoising method of oval search window and parameter adaptive according to claim 1, special
Sign is, in the step (7), by following formula 5 by the R of the pixeliValue and IRFrequency accumulation histogram T in it is maximum
30%, the corresponding threshold value T in 70% and 90% place0.3、T0.7And T0.9It is divided into low noise texture region c1, intermediate region c2, height makes an uproar
Sound texture region c3With flat site c4Four classifications:
6. a kind of non-local mean denoising method of oval search window and parameter adaptive according to claim 1, special
Sign is, in the step (8), with psiIndicate the size of image block at pixel i, different classes of image-region is not using
Same psi;H is calculated by following formula 6i, to indicate smoothing factor values of the pixel i in noise image I:
Wherein,For the design factor of chi square distribution,Indicate degree of freedom
For psiAnd probability value is the chi-square value corresponding to 0.99.
7. a kind of non-local mean denoising method of oval search window and parameter adaptive according to claim 1, special
Sign is, in the step (9), elliptical direction of rotation θ is calculated by following formula 7i:
Wherein, Fi,1(1) and Fi,1(2) structure tensor C is indicated respectivelyiThe feature vector F obtained after decompositioniAlong the directions x and y
Component value;The size of elliptical long axis and short axle is obtained by following formula 8:
Wherein, A and B indicates that the length of transverse and short axle, L are rectangular local neighborhood Ω respectivelyiRadius, σiWith formula (2)
In σiUnanimously.
8. a kind of non-local mean denoising method of oval search window and parameter adaptive according to claim 7, special
Sign is, in the step (9), rectangular local neighborhood Ω is calculated by following formula 9iIn between other pixels and central point i
Coordinate distance:
Wherein, coordinate differences of Δ x and the oval search window of Δ y expressions interior other points and elliptical center point, are obtained by following formula 10
To the elliptic equation E of oval search windowi(z):
It then, can be by the postrotational (x of traversal pointr,yr) coordinate is brought elliptic equation into and judged, if Ei(z)≤1 item it is determined as
Inside ellipse or on ellipse.
9. a kind of non-local mean denoising method of oval search window and parameter adaptive according to claim 1, special
Sign is, in the step (10), the figure in oval search window centered on other pixels j is calculated by following formula 11
As block PjWith the image block P centered on oval centre point iiBetween weighted value Wij:
Wherein, ZiFor normalization factor, hiFor smoothing factor, ΦiFor the oval search window centered on i,Table
Show the weighted euclidean distance between two image blocks.
10. a kind of non-local mean denoising method of oval search window and parameter adaptive according to claim 1, special
Sign is, in the step (12), the image block after denoising is obtained by following formula 12
Finally, abovementioned steps are repeated, the image block after obtaining all pixels point denoising, the pixel value of lap is cumulative to be taken
The image being averagely worth to after denoising.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810502466.XA CN108765332B (en) | 2018-05-23 | 2018-05-23 | Ellipse search window and parameter self-adaptive non-local mean denoising method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810502466.XA CN108765332B (en) | 2018-05-23 | 2018-05-23 | Ellipse search window and parameter self-adaptive non-local mean denoising method |
Publications (2)
Publication Number | Publication Date |
---|---|
CN108765332A true CN108765332A (en) | 2018-11-06 |
CN108765332B CN108765332B (en) | 2022-05-06 |
Family
ID=64005389
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810502466.XA Active CN108765332B (en) | 2018-05-23 | 2018-05-23 | Ellipse search window and parameter self-adaptive non-local mean denoising method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN108765332B (en) |
Cited By (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109785246A (en) * | 2018-12-11 | 2019-05-21 | 深圳奥比中光科技有限公司 | A kind of noise-reduction method of non-local mean filtering, device and equipment |
CN112986964A (en) * | 2021-02-26 | 2021-06-18 | 北京空间机电研究所 | Photon counting laser point cloud self-adaptive denoising method based on noise neighborhood density |
CN115330784A (en) * | 2022-10-13 | 2022-11-11 | 南通金百福纺织品有限公司 | Cloth surface defect detection method |
CN116030055A (en) * | 2023-03-29 | 2023-04-28 | 凯尔菱电(山东)电梯有限公司 | Quality detection method for elevator-related accessories |
CN116168024A (en) * | 2023-04-23 | 2023-05-26 | 苏州大凯纸业有限公司 | Corrugated paper production quality real-time detection method based on computer vision |
CN116205823A (en) * | 2023-05-05 | 2023-06-02 | 青岛市妇女儿童医院(青岛市妇幼保健院、青岛市残疾儿童医疗康复中心、青岛市新生儿疾病筛查中心) | Ultrasonic image denoising method based on spatial domain filtering |
CN116363018A (en) * | 2023-05-30 | 2023-06-30 | 江苏游隼微电子有限公司 | Adaptive NLM noise reduction method |
CN117253144A (en) * | 2023-09-07 | 2023-12-19 | 建研防火科技有限公司 | Fire risk grading management and control method |
CN117496448B (en) * | 2024-01-02 | 2024-04-26 | 深圳市旭仓科技有限公司 | Intelligent monitoring system and method for building construction safety |
Citations (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20070253030A1 (en) * | 2000-09-08 | 2007-11-01 | Kathrin Berkner | Wavelet-based image processing path |
CN101877122A (en) * | 2009-11-12 | 2010-11-03 | 中国科学院遥感应用研究所 | Method for denoising and enhancing anisotropic diffusion image with controllable diffusion degree |
CN102622731A (en) * | 2012-03-11 | 2012-08-01 | 西安电子科技大学 | Contourlet domain Wiener filtering image denoising method based on two-dimensional Otsu |
CN103020918A (en) * | 2013-01-09 | 2013-04-03 | 西安电子科技大学 | Shape-adaptive neighborhood mean value based non-local mean value denoising method |
CN103226808A (en) * | 2012-01-30 | 2013-07-31 | 深圳迈瑞生物医疗电子股份有限公司 | Image enhancement method and image enhancement system |
CN103955903A (en) * | 2014-05-09 | 2014-07-30 | 东南大学 | Weight window self-adaptation non-local mean image denoising method |
CN104599244A (en) * | 2014-12-23 | 2015-05-06 | 中国科学院深圳先进技术研究院 | Denoising method and denoising system for magnetic resonance diffusion tensor imaging |
CN104680495A (en) * | 2015-03-17 | 2015-06-03 | 武汉倍尔生生物科技有限公司 | Adaptive denoising method for ultrasonic image |
CN106023093A (en) * | 2016-05-09 | 2016-10-12 | 吉林大学 | Non-local mean value image denoising method based on improved image black matching |
US20160300328A1 (en) * | 2014-01-17 | 2016-10-13 | Tencent Technology (Shenzhen) Co., Ltd. | Method and apparatus for implementing image denoising |
-
2018
- 2018-05-23 CN CN201810502466.XA patent/CN108765332B/en active Active
Patent Citations (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20070253030A1 (en) * | 2000-09-08 | 2007-11-01 | Kathrin Berkner | Wavelet-based image processing path |
CN101877122A (en) * | 2009-11-12 | 2010-11-03 | 中国科学院遥感应用研究所 | Method for denoising and enhancing anisotropic diffusion image with controllable diffusion degree |
CN103226808A (en) * | 2012-01-30 | 2013-07-31 | 深圳迈瑞生物医疗电子股份有限公司 | Image enhancement method and image enhancement system |
CN102622731A (en) * | 2012-03-11 | 2012-08-01 | 西安电子科技大学 | Contourlet domain Wiener filtering image denoising method based on two-dimensional Otsu |
CN103020918A (en) * | 2013-01-09 | 2013-04-03 | 西安电子科技大学 | Shape-adaptive neighborhood mean value based non-local mean value denoising method |
US20160300328A1 (en) * | 2014-01-17 | 2016-10-13 | Tencent Technology (Shenzhen) Co., Ltd. | Method and apparatus for implementing image denoising |
CN103955903A (en) * | 2014-05-09 | 2014-07-30 | 东南大学 | Weight window self-adaptation non-local mean image denoising method |
CN104599244A (en) * | 2014-12-23 | 2015-05-06 | 中国科学院深圳先进技术研究院 | Denoising method and denoising system for magnetic resonance diffusion tensor imaging |
CN104680495A (en) * | 2015-03-17 | 2015-06-03 | 武汉倍尔生生物科技有限公司 | Adaptive denoising method for ultrasonic image |
CN106023093A (en) * | 2016-05-09 | 2016-10-12 | 吉林大学 | Non-local mean value image denoising method based on improved image black matching |
Non-Patent Citations (2)
Title |
---|
YUMING WANG等: "Text String Extraction from Scene Image Based on Edge Feature and Morphology", 《 2008 THE EIGHTH IAPR INTERNATIONAL WORKSHOP ON DOCUMENT ANALYSIS SYSTEMS》 * |
孙强等: "基于空间自适应Bayesian缩减的NSCT域图像去噪方法", 《计算机应用》 * |
Cited By (15)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109785246A (en) * | 2018-12-11 | 2019-05-21 | 深圳奥比中光科技有限公司 | A kind of noise-reduction method of non-local mean filtering, device and equipment |
CN109785246B (en) * | 2018-12-11 | 2021-04-30 | 奥比中光科技集团股份有限公司 | Noise reduction method, device and equipment for non-local mean filtering |
CN112986964A (en) * | 2021-02-26 | 2021-06-18 | 北京空间机电研究所 | Photon counting laser point cloud self-adaptive denoising method based on noise neighborhood density |
CN112986964B (en) * | 2021-02-26 | 2023-03-31 | 北京空间机电研究所 | Photon counting laser point cloud self-adaptive denoising method based on noise neighborhood density |
CN115330784A (en) * | 2022-10-13 | 2022-11-11 | 南通金百福纺织品有限公司 | Cloth surface defect detection method |
CN116030055B (en) * | 2023-03-29 | 2023-07-11 | 凯尔菱电(山东)电梯有限公司 | Quality detection method for elevator-related accessories |
CN116030055A (en) * | 2023-03-29 | 2023-04-28 | 凯尔菱电(山东)电梯有限公司 | Quality detection method for elevator-related accessories |
CN116168024A (en) * | 2023-04-23 | 2023-05-26 | 苏州大凯纸业有限公司 | Corrugated paper production quality real-time detection method based on computer vision |
CN116168024B (en) * | 2023-04-23 | 2023-07-14 | 苏州大凯纸业有限公司 | Corrugated paper production quality real-time detection method based on computer vision |
CN116205823A (en) * | 2023-05-05 | 2023-06-02 | 青岛市妇女儿童医院(青岛市妇幼保健院、青岛市残疾儿童医疗康复中心、青岛市新生儿疾病筛查中心) | Ultrasonic image denoising method based on spatial domain filtering |
CN116363018A (en) * | 2023-05-30 | 2023-06-30 | 江苏游隼微电子有限公司 | Adaptive NLM noise reduction method |
CN116363018B (en) * | 2023-05-30 | 2023-08-08 | 江苏游隼微电子有限公司 | Adaptive NLM noise reduction method |
CN117253144A (en) * | 2023-09-07 | 2023-12-19 | 建研防火科技有限公司 | Fire risk grading management and control method |
CN117253144B (en) * | 2023-09-07 | 2024-04-12 | 建研防火科技有限公司 | Fire risk grading management and control method |
CN117496448B (en) * | 2024-01-02 | 2024-04-26 | 深圳市旭仓科技有限公司 | Intelligent monitoring system and method for building construction safety |
Also Published As
Publication number | Publication date |
---|---|
CN108765332B (en) | 2022-05-06 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN108765332A (en) | A kind of non-local mean denoising method of ellipse search window and parameter adaptive | |
Agrawal et al. | A novel joint histogram equalization based image contrast enhancement | |
CN115861135B (en) | Image enhancement and recognition method applied to panoramic detection of box body | |
CN107392866B (en) | A kind of facial image local grain Enhancement Method of illumination robust | |
CN113963042B (en) | Metal part defect degree evaluation method based on image processing | |
CN108022233A (en) | A kind of edge of work extracting method based on modified Canny operators | |
Chabrier et al. | Unsupervised evaluation of image segmentation application to multi-spectral images | |
CN102298773B (en) | Shape-adaptive non-local mean denoising method | |
CN108921800A (en) | Non-local mean denoising method based on form adaptive search window | |
CN107506795A (en) | A kind of local gray level histogram feature towards images match describes sub- method for building up and image matching method | |
CN111161222B (en) | Printing roller defect detection method based on visual saliency | |
CN109064418A (en) | A kind of Images Corrupted by Non-uniform Noise denoising method based on non-local mean | |
CN106991661B (en) | Non-local mean denoising method fusing KL (karhunen-Loeve) transformation and grey correlation degree | |
CN111832642A (en) | Image identification method based on VGG16 in insect taxonomy | |
CN111882555B (en) | Deep learning-based netting detection method, device, equipment and storage medium | |
CN108510478B (en) | Lung airway image segmentation method, terminal and storage medium | |
CN110599422B (en) | Weighted average salt and pepper noise reduction method based on edge protection | |
CN115272303A (en) | Textile fabric defect degree evaluation method, device and system based on Gaussian blur | |
CN111179295A (en) | Improved two-dimensional Otsu threshold image segmentation method and system | |
CN113378620A (en) | Cross-camera pedestrian re-identification method in surveillance video noise environment | |
CN110910347B (en) | Tone mapping image non-reference quality evaluation method based on image segmentation | |
CN114359083B (en) | High-dynamic thermal infrared image self-adaptive preprocessing method for interference environment | |
CN110889806A (en) | Adaptive gain image enhancement method based on fractional order multi-scale entropy fusion | |
CN113759375B (en) | SAR image non-local mean denoising method based on statistical characteristics | |
CN107085832A (en) | A kind of Fast implementation of the non local average denoising of digital picture |
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 |