CN102663706A - Adaptive weighted mean value filtering method based on diamond template - Google Patents
Adaptive weighted mean value filtering method based on diamond template Download PDFInfo
- Publication number
- CN102663706A CN102663706A CN2012101198650A CN201210119865A CN102663706A CN 102663706 A CN102663706 A CN 102663706A CN 2012101198650 A CN2012101198650 A CN 2012101198650A CN 201210119865 A CN201210119865 A CN 201210119865A CN 102663706 A CN102663706 A CN 102663706A
- Authority
- CN
- China
- Prior art keywords
- pixel
- template
- rhombus
- noise
- pixels
- 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.)
- Pending
Links
Images
Landscapes
- Image Processing (AREA)
Abstract
The invention relates to an adaptive weighted mean value filtering method based on a diamond template. In the method, impulse noise is removed by weighting images to obtain a mean value according to the overlapped area of signal pixels in a center pixel neighborhood through the diamond template. During noise removal, the diamond template is adaptively extended according to the difference of noise densities and the need on the number of signal pixels in the diamond template, the number of signal points involved in operation when the size of the diamond template is nXn (n is an odd number larger than or equal to 5) is smaller than that of pixel points involved in operation when a square template is used, the larger the diamond template is, the more obvious the difference is, and accordingly, noise removal speed can be increased for the images with different noise densities. In addition, as only the gray value of the signal pixels in the diamond template is allowed to be involved in calculation of the gray value of center pixels, a filtering effect is obviously improved. The images polluted by impulse noise with various densities can be effectively filtered.
Description
Technical field
The present invention relates to a kind of adaptive weighted mean filter method, be specially adapted to filtering image salt-pepper noise, belong to technical field of image processing based on the rhombus template.
Background technology
Growing and universal along with computing machine and all kinds of electronic imaging devices; Digital picture with the closely bound up every field of people's social life in obtain application more and more widely, the remote sensing SAR image of taking like the used CT image of: medical science, spacecraft, the digital watermarking image of the checking bank note check true and false, people's identity is known others face iris image or the like.This will ask for help can handle more accurately image, thereby extract and analyze more information and characteristic effectively.
Noise is one of principal element that influences picture quality and visual effect; The main cause of its generation be image obtain with transmission course in because image capturing system, transmission medium and the imperfection of imaging system and the interference of external environment make picture quality receive infringement.Thereby noise hinders people's acceptance pattern as the original information influence advanced processes follow-up to image, like rim detection, image segmentation, feature extraction, pattern-recognition etc.So how to study filtering image noise, improve picture quality just become in the digital picture research field one extremely important and have a problem of realistic meaning.
Image denoising is exactly to use effective filtering method the noise that is contained in the image is removed as much as possible, promptly the information of carrying out of the information defect area on the image is filled, and makes filtered image energy farthest near original image.
At present more common filtering template have n * n square (n for more than or equal to 3 odd number) usually template, cruciform template, * shape template, all directions be to template etc.Wherein square template applications is the most extensive, and when handling some pixels, the pixel of getting in its n * n square template is participated in filtering calculating.It has utilized a large amount of Pixel Information around this pixel, aspect program design, also realizes than being easier to, and can obtain reasonable effect under a lot of situation.
Traditional filtering method is medium filtering and mean filter.Medium filtering is a non-linear filtering method; It also is preconditioning technique the most frequently used in the image processing techniques; Basic thought is: for each pixel in the given image; To be that the gray-scale value of all pixels in the n * n square template at center sorts with it, get the final gray-scale value of intermediate value then as this pixel.This method can overcome linear filter blurs to what image brought, in effective eliminating particle noise, can keep good local edge again, thereby obtain satisfied filter effect, is particularly suitable for removing the salt-pepper noise of image; Mean filter is the linear filtering algorithm, is neighborhood averaging again, and basic thought is: for each pixel in the given image, getting with it is the center
The mean value of all grey scale pixel values is as the gray-scale value of current pixel in n * n square template, and this method computing is simple, and Gaussian noise is had the good denoising ability.
Though use n * n square template filtering to utilize the half-tone information of the most of pixel around the current pixel, it does not give full play to all half-tone informations and the effect of range information in filtering in the current pixel neighborhood.
Summary of the invention
Technical matters to be solved by this invention provides a kind of adaptive weighted mean filter method based on the rhombus template.
The technical scheme that technical solution problem of the present invention is adopted:
The present invention includes following steps:
(1) reads in piece image, make that first pixel is a current pixel;
(2) judge whether current pixel is noise:
The method of judging noise is the method that extreme value and threshold interval combine; Even the gray-scale value of current pixel is the maximum value or minimum value of all grey scale pixel values in its 3 * 3 rhombus template neighborhood; And its gray-scale value is when [251,255] or [0,4]; Judge that this pixel is a noise, otherwise be normal signal; Said noise is referred to as noise pixel, and said normal signal is referred to as signal pixels;
Center symmetry and rotational symmetry figure that said rhombus template is meant is that catercorner length equates, diagonal line is respectively horizontal direction and vertical direction; Said 3 * 3 rhombus templates are meant that the number of pixels that two diagonal line relate to all is 3, being that four points in upper and lower, left and right of 3/2 pixel distance are the said rhombus template of summit structure apart from the rhombus center;
When said current pixel is noise pixel, carried out for (3) step;
When said current pixel is signal pixels, carried out for (2.1) step;
(2.1) putting next pixel is current pixel, returns for (2) step then;
(3) with the noise pixel be the center, construct said 3 * 3 rhombus templates:
When the pixel with image boundary was central configuration rhombus template, the part that exceeds image range in the rhombus template was not considered, and miscount takes place when preventing to be beyond the boundary;
(4) >=2 whether the number of signal pixels in the decision diamond template:
When the number of signal pixels in the rhombus template >=2, carried out for (5) step;
When the number of signal pixels in the rhombus template less than 2 the time, carried out for (4.1) step;
(4.1) the rhombus template expands:
The extending method of rhombus template is following:
During each the expansion, be that the center respectively increases by 2 with the pixel on two diagonal line with the current pixel, constitute new rhombus template, expand successively, until extending to n * n rhombus template, wherein n is an odd number;
Returning after the each expansion of rhombus template finishes changeed for (4) step;
(5) press the area weight P that following formula (1) calculates each signal pixels in the rhombus template
i,
P
i=S
i/S （1）
In the formula, P
iIt is the area weight of i signal pixels;
S
iArea for i signal pixels in the rhombus template and rhombus template lap;
S is the area of rhombus template;
In calculating, to establish each pixel and be square, its area is 1; And other noise pixel does not participate in calculating in center pixel and the rhombus template;
(6) according to the following equation the normalization weight value W that each participates in the signal pixels of calculating is calculated in (2)
i:
W
i=P
i/∑P
i （2）
In the formula, W
iIt is the normalization weight value of i signal pixels;
(7) remove noise, promptly calculate weighted mean C, and with the gray-scale value of this weighted mean C as center pixel, said center pixel be a current pixel by following formula (3):
C=∑W
i*C
i （3）
In the formula, W
iBe the normalization weight value of i signal pixels, C
iGray-scale value for i signal pixels in the rhombus template;
(8) judge whether all pixels dispose:
When all pixels do not dispose, returned for (2.1) step;
When all processes pixel finish, got into for (9) step;
(9) denoising finishes.
The present invention utilized the peripheral information of current pixel to come filtering noise fully, and the big I of rhombus template expands adaptively through adopting n * n rhombus template according to noise density.When calculating, through the area weighting, make that the information role of the signal pixels near apart from current pixel is big, the information role of the signal pixels far away apart from current pixel is little, has improved filter effect.
For the image that polluted by salt-pepper noise, as pending pixel, be the center with noise, according to the density of the noise symmetrical expansion templates that outwards assumes diamond in shape adaptively with pending pixel.Utilize in this template signal pixels and the overlapping area of template to carry out the weighting calculating of averaging, obtain the gray-scale value of pending pixel.The gray-scale value of other noise pixel is not participated in the calculating of center noise pixel gray-scale value in the rhombus template; Total energy guarantees to have the gray-scale value of enough signal pixels to participate in weighted mean calculating; Use in addition with the gray-scale value of the pending pixel of information calculations of the signal pixels of rhombus template lap area non-zero than normally used square template more rationally, effectively, can effectively reduce calculated amount in the denoising.
Beneficial effect of the present invention is following:
(1) advantage of employing rhombus template:
In a, the rhombus template, the weight of participating in the signal pixels of calculating is associated with the size of rhombus template overlapping area with it, has spatially embodied the Different Effects of the signal pixels of diverse location to center pixel.
The diagonal line of b, rhombus template equates to have good symmetry up and down, even therefore template constantly enlarges, the pixel of participating in computing constantly increases, and its weighted value also is simple and easy to calculate, and has significantly reduced computing time and complexity.
C, compare with square template; All signal pixels of participating in calculating in the rhombus template are less to the distance variance of central point; Avoid the pixel of decentering point hypertelorism to participate in filtering calculating as far as possible, meet generally speaking the strong more actual conditions of the nearlyer gray-scale value correlativity of distance more.
D, when template is 3 * 3, the rhombus template is the same with the number of pixels (comprising signal pixels and noise pixel) that square template relates to.But along with the expansion of template, scope that the rhombus template relates to and number of pixels are less than scope and the number of pixels that square template relates to, and like this, have both reduced calculated amount, have reflected the influence degree of different pixels to current pixel again scientifically and rationally.
E, in filtering, consider signal pixels (non-noise pixel) number in the rhombus template.Number very little, filter effect is bad naturally, when the number of signal pixels in the rhombus template does not reach minimum value 2, expands the rhombus template automatically, has adaptivity.
F, experimental result show that the present invention can handle the low-density noise, also can handle the high density noise, and noise density is high more, and through the self-adaptation expansion of template, denoising effect is good more relatively.
If the pixel of g image boundary is a noise, then for being that the part that exceeds image-region in the rhombus template of central configuration is not considered the miscount that has taken place when effectively having prevented to be beyond the boundary with this pixel.
(2) the present invention only allows the signal pixels in the rhombus template to participate in the calculating of current pixel gray-scale value, effectively reduces the influence of other noise on filtering effects that possibly exist in the rhombus template, makes filter effect more near original image.
(3) filtering of rhombus template and n * n square template filtering is compared, the filtered signal to noise ratio (S/N ratio) of rhombus template be largely increased (seeing attached list 1).
Description of drawings
Fig. 1 is an algorithm flow chart of the present invention.
Fig. 2 (a)-2 (c) is respectively 3 * 3,5 * 5,7 * 7 rhombus template figure.
Fig. 3 (a)-3 (c) is respectively the area weight distribution figure of each pixel in 3 * 3,5 * 5,7 * 7 the rhombus template.
Fig. 4 (a) is original Lena figure.
Fig. 4 (b)-4 (k) is respectively to add and makes an uproar 4%, 10%, 20%, 30%, 40%, 50%, 60%, 70%, 80%, 90% Lena figure.
Fig. 4 (b1)-4 (k1) is respectively and uses 3 * 3 square templates that the Lena figure of above-mentioned different noise densities is carried out the design sketch behind the medium filtering.
Fig. 4 (b2)-4 (k2) is respectively and uses rhombus template of the present invention that the Lena figure of above-mentioned different noise densities is carried out the design sketch after the weighted filtering.
Embodiment
As shown in Figure 1, the present invention includes following steps:
(1) reads in piece image, make that first pixel is a current pixel;
(2) judge whether current pixel is noise:
The method of judging noise is the method that extreme value and threshold interval combine; Even the gray-scale value of current pixel is the maximum value or minimum value of all grey scale pixel values in its 3 * 3 rhombus template neighborhood; And its gray-scale value is when [251,255] or [0,4]; Judge that this pixel is a noise, otherwise be normal signal; Said noise is referred to as noise pixel, and said normal signal is referred to as signal pixels;
Center symmetry and rotational symmetry figure that said rhombus template is meant is that catercorner length equates, diagonal line is respectively horizontal direction and vertical direction; Said 3 * 3 rhombus templates are meant that the number of pixels that two diagonal line relate to all is 3, being that four points in upper and lower, left and right of 3/2 pixel distance are the said rhombus template of summit structure apart from the rhombus center;
When said current pixel is noise pixel, carried out for (3) step;
When said current pixel is signal pixels, carried out for (2.1) step;
(2.1) putting next pixel is current pixel, returns for (2) step then;
(3) with the noise pixel be the center, construct said 3 * 3 rhombus templates:
When the pixel with image boundary was central configuration rhombus template, the part that exceeds image range in the rhombus template was not considered, and miscount takes place when preventing to be beyond the boundary;
(4) >=2 whether the number of signal pixels in the decision diamond template:
When the number of signal pixels in the rhombus template >=2, carried out for (5) step;
When the number of signal pixels in the rhombus template less than 2 the time, carried out for (4.1) step;
(4.1) the rhombus template expands:
The extending method of rhombus template is following:
During each the expansion; With the current pixel is that the center respectively increases by 2 with the pixel on two diagonal line, constitutes new rhombus template, expands successively; Until extending to n * n rhombus template; Wherein n is that (referring to Fig. 2 (a)-2 (c), when the current pixel, current pixel is positioned at rhombus template center to odd number with the rhombus template action; It waits all directions to have about central point center symmetry with about the axisymmetric character of diagonal line up and down.Therefore; In the rhombus template, equate about central point center symmetry with about the area of the axisymmetric pixel of diagonal line with rhombus template lap, not that the center is symmetrical and be not that the area of axisymmetric pixel and rhombus template lap is unequal about diagonal line about central point; And pixel decentering pixel is near more; The area of it and rhombus template lap is big more, and pixel decentering pixel is far away more, and the area of it and rhombus template lap is more little.During the gray-scale value of computing center's pixel, confirm the area weight of different pixels in the rhombus template according to the size of lap area.
Returning after the each expansion of rhombus template finishes changeed for (4) step;
(5) press the area weight P that following formula (1) calculates each signal pixels in the rhombus template
i,
P
i=S
i/S （1）
In the formula, P
iIt is the area weight of i signal pixels;
S
iArea for i signal pixels in the rhombus template and rhombus template lap;
S is the area of rhombus template;
In calculating, to establish each pixel and be square, its area is 1; And other noise pixel does not participate in calculating in center pixel and the rhombus template;
(6) according to the following equation the normalization weight value W that each participates in the signal pixels of calculating is calculated in (2)
i:
W
i=P
i/∑P
i （2）
In the formula, W
iIt is the normalization weight value of i signal pixels;
(7) remove noise, promptly calculate weighted mean C, and with the gray-scale value of this weighted mean C as center pixel, said center pixel be a current pixel by following formula (3):
C=∑W
i*C
i （3）
In the formula, W
iBe the normalization weight value of i signal pixels, C
iGray-scale value for i signal pixels in the rhombus template;
(8) judge whether all pixels dispose:
When all pixels do not dispose, returned for (2.1) step;
When all processes pixel finish, got into for (9) step;
(9) denoising finishes.
Below in conjunction with accompanying drawing present embodiment is further specified:
Fig. 2 (a)-2 (c) is respectively 3 * 3,5 * 5,7 * 7 rhombus templates, and (i is current pending noise j) to pixel wherein, is positioned at the center of each rhombus template.Fig. 2 (a) is an original template, and the number of pixels on each diagonal line is 3, is referred to as 3 * 3 rhombus templates; Fig. 2 (b) is the rhombus template after expanding for the first time, and the number of pixels on each diagonal line is 5, is referred to as 5 * 5 rhombus templates; Fig. 2 (c) is 7 * 7 rhombus templates after expanding, and the number of pixels on each diagonal line is 7.Similar expansion can get bigger n * n (n is not less than 5 odd number) rhombus template, and the number of pixels on each diagonal line is n.
Fig. 3 (a)-3 (c) is respectively the area weight calculation example of each pixel in 3 * 3,5 * 5,7 * 7 rhombus templates.(i j) is current noise to the center pixel of rhombus template, and in the self-adaptation expansion process, the rhombus template action that varies in size is when current pixel, and the pixel in the rhombus template and the size of template lap area are shown in Fig. 3 (a)-3 (c).If each pixel is a unit square, its area is 1.In 3 * 3 rhombus templates of Fig. 3 (a); Pixel in the rhombus template has 8 (not comprise pending center pixel; Other rhombus template also is like this); Four pixels in the upper and lower, left and right of center pixel and the overlapping area of rhombus template are 3/4, it is upper left, upper right, a left side down, four pixels in bottom right are 1/8 with the overlapping area of rhombus template, the area of rhombus template is 7/2 at this moment; After the normalization, the area weighted value of four pixels in upper and lower, left and right of center pixel is P
i=(3/4)/(7/2)=3/14, upper left, upper right a, left side is descended, the area weighted value P of four pixels in bottom right
i=(1/8)/(7/2)=1/28;
In 5 * 5 rhombus templates of Fig. 3 (b), the pixel in the rhombus template has 20, and they and the overlapping area of rhombus template are respectively 1/8,3/4,7/8,1, and this moment, the area of rhombus template was 23/2, after the normalization, and the area weighted value P of these pixels
iBe respectively 1/92,3/46,7/92,2/23;
In 7 * 7 rhombus templates of Fig. 3 (c), the pixel in the rhombus template has 36, and they and the overlapping area of rhombus template are respectively 1/8,3/4,7/8,1, and this moment, the area of rhombus template was 47/2, after the normalization, and the area weighted value P of these pixels
iBe respectively 1/188,3/94,7/188,2/47.
Utilize formula W
i=P
i/ ∑ P
iCalculate the normalization weight W that each participates in the signal pixels of calculating
i, P wherein
iBe the area weighted value of i signal pixels in the current rhombus template, utilize formula C=∑ W at last
i* C
iComputing center's pixel is the gray-scale value C of current noise pixel, wherein C
iGray-scale value for i signal pixels in the current rhombus template.
Fig. 4 (a)-4 (k2) is the filter effect comparison diagrams of different filtering methods to the different densities noise image; In theory, the rhombus template can infinitely expand.But in the present embodiment, only let the rhombus template extend to 7 * 7;
Fig. 4 (a) is original Lena figure; Fig. 4 (b)-4 (k) is respectively to add and makes an uproar 4%, 10%, 20%, 30%, 40%, 50%, 60%, 70%, 80%, 90% Lena figure;
Fig. 4 (b1)-4 (k1) carries out the design sketch behind the medium filtering for using 3 * 3 square templates to the Lena figure of above-mentioned different noise densities;
Fig. 4 (b2)-4 (k2) carries out the design sketch after the weighted filtering for using the rhombus template to the Lena figure of above-mentioned different noise densities.
Claims (1)
1. adaptive weighted mean filter method based on the rhombus template is characterized in that may further comprise the steps:
(1) reads in piece image, make that first pixel is a current pixel;
(2) judge whether current pixel is noise:
The method of judging noise is the method that extreme value and threshold interval combine; Even the gray-scale value of current pixel is the maximum value or minimum value of all grey scale pixel values in its 3 * 3 rhombus template neighborhood; And its gray-scale value is when [251,255] or [0,4]; Judge that this pixel is a noise, otherwise be normal signal; Said noise is referred to as noise pixel, and said normal signal is referred to as signal pixels;
Center symmetry and rotational symmetry figure that said rhombus template is meant is that catercorner length equates, diagonal line is respectively horizontal direction and vertical direction; Said 3 * 3 rhombus templates are meant that the number of pixels that two diagonal line relate to all is 3, being that four points in upper and lower, left and right of 3/2 pixel distance are the said rhombus template of summit structure apart from the rhombus center;
When said current pixel is noise pixel, carried out for (3) step;
When said current pixel is signal pixels, carried out for (2.1) step;
(2.1) putting next pixel is current pixel, returns for (2) step then;
(3) with the noise pixel be the center, construct said 3 * 3 rhombus templates:
When the pixel with image boundary was central configuration rhombus template, the part that exceeds image range in the rhombus template was not considered, and miscount takes place when preventing to be beyond the boundary;
(4) >=2 whether the number of signal pixels in the decision diamond template:
When the number of signal pixels in the rhombus template >=2, carried out for (5) step;
When the number of signal pixels in the rhombus template less than 2 the time, carried out for (4.1) step;
(4.1) the rhombus template expands:
The extending method of rhombus template is following:
During each the expansion, be that the center respectively increases by 2 with the pixel on two diagonal line with the current pixel, constitute new rhombus template, expand successively, until extending to n * n rhombus template, wherein n is an odd number;
Returning after the each expansion of rhombus template finishes changeed for (4) step;
(5) press the area weight P that following formula (1) calculates each signal pixels in the rhombus template
i,
P
i=S
i/S （1）
In the formula, P
iIt is the area weight of i signal pixels;
S
iArea for i signal pixels in the rhombus template and rhombus template lap;
S is the area of rhombus template;
In calculating, to establish each pixel and be square, its area is 1; And other noise pixel does not participate in calculating in center pixel and the rhombus template;
(6) according to the following equation the normalization weight value W that each participates in the signal pixels of calculating is calculated in (2)
i:
W
i=P
i/∑P
i （2）
In the formula, W
iIt is the normalization weight value of i signal pixels;
(7) remove noise, promptly calculate weighted mean C, and with the gray-scale value of this weighted mean C as center pixel, said center pixel be a current pixel by following formula (3):
C=∑W
i*C
i （3）
In the formula, W
iBe the normalization weight value of i signal pixels, C
iGray-scale value for i signal pixels in the rhombus template;
(8) judge whether all pixels dispose:
When all pixels do not dispose, returned for (2.1) step;
When all processes pixel finish, got into for (9) step;
(9) denoising finishes.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2012101198650A CN102663706A (en) | 2012-04-23 | 2012-04-23 | Adaptive weighted mean value filtering method based on diamond template |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2012101198650A CN102663706A (en) | 2012-04-23 | 2012-04-23 | Adaptive weighted mean value filtering method based on diamond template |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN102663706A true CN102663706A (en) | 2012-09-12 |
Family
ID=46773185
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN2012101198650A Pending CN102663706A (en) | 2012-04-23 | 2012-04-23 | Adaptive weighted mean value filtering method based on diamond template |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN102663706A (en) |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103258318A (en) * | 2013-05-24 | 2013-08-21 | 苏州科达科技股份有限公司 | Method and system for image noise reduction processing |
| CN103400357A (en) * | 2013-08-23 | 2013-11-20 | 闽江学院 | Method for removing salt-pepper noises in images |
| CN103761707A (en) * | 2013-12-20 | 2014-04-30 | 浙江大学 | Average filtering method eliminating image impulse noise fast and efficiently |
| CN104240185A (en) * | 2013-06-08 | 2014-12-24 | 深圳深讯和科技有限公司 | Image filtering method and device |
| CN105931197A (en) * | 2016-04-13 | 2016-09-07 | 石家庄铁道大学 | Image de-noising method based on ambiguity theory |
| CN109614855A (en) * | 2018-10-31 | 2019-04-12 | 温州医科大学 | The lagophthalmos After Cataract analytical equipment and After Cataract light and heavy degree evaluation method of analysis are calculated based on gray value of image |
| CN112561823A (en) * | 2020-12-18 | 2021-03-26 | 上海国茂数字技术有限公司 | Filtering method |
| CN115294097A (en) * | 2022-09-26 | 2022-11-04 | 南通君与晴纺织有限公司 | Textile surface defect detection method based on machine vision |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6535860B1 (en) * | 1999-08-02 | 2003-03-18 | National Science Council | Design and hardware synthesis of adaptive weighted fuzzy mean image filter |
| CN102256048A (en) * | 2011-07-19 | 2011-11-23 | 南京信息工程大学 | Density-adaptive image salt-pepper noise switching filtering method |
-
2012
- 2012-04-23 CN CN2012101198650A patent/CN102663706A/en active Pending
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6535860B1 (en) * | 1999-08-02 | 2003-03-18 | National Science Council | Design and hardware synthesis of adaptive weighted fuzzy mean image filter |
| CN102256048A (en) * | 2011-07-19 | 2011-11-23 | 南京信息工程大学 | Density-adaptive image salt-pepper noise switching filtering method |
Non-Patent Citations (5)
| Title |
|---|
| 何蕾等: "一种线性与非线性相结合的图像缩小方法", 《计算机应用研究》 * |
| 唐彩虹等: "《一种基于直方图的加权均值滤波方法》", 《微计算机信息》 * |
| 王晓凯等: "《改进的自适应中值滤波》", 《计算机工程与应用》 * |
| 穆望舒等: "《加权SUSAN算法与遥感图像矢量化》", 《计算机光盘软件与应用》 * |
| 郭明等: "《去除椒盐噪声的非对称有向窗加权均值滤波》", 《激光与红外》 * |
Cited By (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103258318A (en) * | 2013-05-24 | 2013-08-21 | 苏州科达科技股份有限公司 | Method and system for image noise reduction processing |
| CN103258318B (en) * | 2013-05-24 | 2015-09-16 | 苏州科达科技股份有限公司 | A kind of image noise reduction disposal route and system |
| CN104240185A (en) * | 2013-06-08 | 2014-12-24 | 深圳深讯和科技有限公司 | Image filtering method and device |
| CN103400357B (en) * | 2013-08-23 | 2016-02-17 | 闽江学院 | A kind of method removing image salt-pepper noise |
| CN103400357A (en) * | 2013-08-23 | 2013-11-20 | 闽江学院 | Method for removing salt-pepper noises in images |
| CN103761707B (en) * | 2013-12-20 | 2016-04-13 | 浙江大学 | A kind of mean filter method rapidly and efficiently removing image salt-pepper noise |
| CN103761707A (en) * | 2013-12-20 | 2014-04-30 | 浙江大学 | Average filtering method eliminating image impulse noise fast and efficiently |
| CN105931197A (en) * | 2016-04-13 | 2016-09-07 | 石家庄铁道大学 | Image de-noising method based on ambiguity theory |
| CN109614855A (en) * | 2018-10-31 | 2019-04-12 | 温州医科大学 | The lagophthalmos After Cataract analytical equipment and After Cataract light and heavy degree evaluation method of analysis are calculated based on gray value of image |
| CN109614855B (en) * | 2018-10-31 | 2023-04-07 | 温州医科大学 | Post cataract analysis device and method based on image gray value calculation and analysis |
| CN112561823A (en) * | 2020-12-18 | 2021-03-26 | 上海国茂数字技术有限公司 | Filtering method |
| CN112561823B (en) * | 2020-12-18 | 2024-05-28 | 上海国茂数字技术有限公司 | Filtering method |
| CN115294097A (en) * | 2022-09-26 | 2022-11-04 | 南通君与晴纺织有限公司 | Textile surface defect detection method based on machine vision |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN102663706A (en) | Adaptive weighted mean value filtering method based on diamond template | |
| Xuan et al. | An improved canny edge detection algorithm | |
| CN105787902B (en) | Utilize the image denoising method of block sorting detection noise | |
| CN104657945B (en) | The infrared small target detection method of multiple dimensioned space-time Federated filter under complex background | |
| CN102156996B (en) | Image edge detection method | |
| CN108022233A (en) | A kind of edge of work extracting method based on modified Canny operators | |
| CN101930597B (en) | Mathematical morphology-based image edge detection method | |
| CN106910169A (en) | A kind of image salt-pepper noise minimizing technology for preventing edge blurry | |
| TWI393073B (en) | Image denoising method | |
| CN105809642B (en) | A method of the natural image based on L0 regularizations is blind to remove motion blur | |
| CN101853497A (en) | Image enhancement method and device | |
| CN104463814B (en) | Image enhancement method based on local texture directionality | |
| CN103198455B (en) | A kind of image de-noising method utilizing total variation minimization and gray level co-occurrence matrixes | |
| Zhang et al. | Decision-based non-local means filter for removing impulse noise from digital images | |
| CN106910170B (en) | A kind of minimizing technology of image salt-pepper noise | |
| CN103208104B (en) | A kind of image de-noising method based on nonlocal theory | |
| CN109377450A (en) | A kind of edge-protected denoising method | |
| CN103020959B (en) | Gravity model-based oceanic front information extraction method | |
| Feng et al. | A new edge detection algorithm based on Canny idea | |
| Fan et al. | Moving target detection algorithm based on Susan edge detection and frame difference | |
| CN105608689B (en) | A kind of panoramic mosaic elimination characteristics of image error hiding method and device | |
| CN103093443A (en) | Image impulse noise self-adaptive filtering method based on genetic algorithm-back propagation (GA-BP) neural network | |
| CN105590301A (en) | Impulse noise elimination method of self-adaption normal-inclined double cross window mean filtering | |
| CN105469428A (en) | Morphological filtering and SVD (singular value decomposition)-based weak target detection method | |
| CN108090492B (en) | Contour detection method based on scale clue suppression |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C12 | Rejection of a patent application after its publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20120912 |