CN103150733A - Self-adapting multi-stage weighted median filtering algorithm applied to digital images - Google Patents
Self-adapting multi-stage weighted median filtering algorithm applied to digital images Download PDFInfo
- Publication number
- CN103150733A CN103150733A CN2013100973637A CN201310097363A CN103150733A CN 103150733 A CN103150733 A CN 103150733A CN 2013100973637 A CN2013100973637 A CN 2013100973637A CN 201310097363 A CN201310097363 A CN 201310097363A CN 103150733 A CN103150733 A CN 103150733A
- Authority
- CN
- China
- Prior art keywords
- value
- point
- noise
- noise spot
- filtering algorithm
- 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
- 238000001914 filtration Methods 0.000 title claims abstract description 48
- 238000000034 method Methods 0.000 claims abstract description 35
- 238000004364 calculation method Methods 0.000 claims abstract description 5
- 230000003044 adaptive effect Effects 0.000 claims description 19
- 230000011664 signaling Effects 0.000 claims description 13
- 230000006978 adaptation Effects 0.000 claims 1
- 238000001514 detection method Methods 0.000 abstract description 13
- 230000000694 effects Effects 0.000 description 6
- 239000006002 Pepper Substances 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 238000009826 distribution Methods 0.000 description 1
- 230000008030 elimination Effects 0.000 description 1
- 238000003379 elimination reaction Methods 0.000 description 1
- 238000012163 sequencing technique Methods 0.000 description 1
Images
Abstract
The invention relates to a self-adapting multi-stage weighted median filtering algorithm applied to digital images. According to the method, the noise point in the image is judged according to the characteristics of the noise value, in addition, the noise point is marked, then, a self-adapting multi-stage weighted median filtering method is adopted for carrying out filtering processing on the noise point, and in addition, in the filtering algorithm realization process, the detected noise point and the treated noise point do not take part in the calculation. The self-adapting multi-stage weighted median filtering algorithm has the advantages that the noise detection correctness is high, the similarity of the processed pixel point to the original image signal is high, and the image detail remaining capability is good.
Description
Technical field
The present invention relates to a kind of filtering algorithm of digital picture, specifically, relate to a kind of adaptive multistage weighted median filtering algorithm that is applied to digital picture.
Background technology
Median filtering algorithm mainly contains at present: conventional median filtering algorithm, extreme value median filtering algorithm, weighted median filtering algorithm, adaptive median filter algorithm and modified median filtering algorithm.Conventional median filtering algorithm is based on a kind of nonlinear signal processing technology that can effectively suppress noise of sequencing statistical theory, the ultimate principle of conventional medium filtering is that the value of any in digital picture or Serial No. is replaced with the Mesophyticum of each point value in a neighborhood of this point, the actual value that pixel value around allowing approaches, thus isolated noise spot eliminated.The extreme value median filtering algorithm is a bit to carry out window operation with certain in image, namely centered by certain point, by a certain window shape such as rectangle, cruciform etc., get several pixels, if the gray-scale value of this point is maximum or minimum value in its neighborhood, this is noise spot so.Otherwise, be signaling point, then noise spot is carried out medium filtering and process.The weighted median filtering algorithm is that the first using right value multiplies each other with a plurality of pixel values respectively and regulates, and then obtains intermediate value in all products.The adaptive median filter algorithm is operated on two levels, A layer and B layer, and arthmetic statement is as follows:
Set up an office (x, y) for given window S
xyThe center, wherein:
Z
min=S
xyIn minimum gradation value
Z
max=S
xyIn maximum gradation value
Z
med=S
xyIn the intermediate value of gray-scale value
Z
xyThe gray-scale value that=coordinate (x, y) is located
S
max=S
xyThe full-size that allows
The adaptive median filter algorithm is expressed as process A and process B with two process work, and is as follows:
Process A:A
1=Z
med-Z
min
A
2＝Z
med-Z
max
If A
1＞0 and A
2＜0, forward process B to
Otherwise increase window size
If window size≤S
max, process repeats A
Otherwise output Z
med
Process B:B
1=Z
xy-Z
min
B
2＝Z
xy-Z
max
If Z is exported in B1＞0 and B2＜0
xy
Otherwise output Z
med
Adaptive median filter is first sought salt-pepper noise at the A layer by algorithm, and the words algorithm that finds can increase window size and repeat, until find a non-pulse intermediate value or reach maximum window size.
The modified median filtering algorithm first carries out the detection of noise spot, according to the characteristic distributions of salt-pepper noise, directly defines the noise spot scope, or carries out the secondary detection of noise spot, then in conjunction with Multilevel Median Filtering Algorithm or the weighted median filtering algorithm carries out filtering.
Conventional median filtering algorithm is all done intermediate value to all pixels in image and is processed, although can the filtering noise point, also changed signaling points a large amount of in the image, reduced the output image signal to noise ratio (S/N ratio), caused the fuzzy of image.The denoising effect of conventional medium filtering is too dependent on the size of filter window and participates in the number of the pixel of median calculation.The selection of filter window is difficult, if filter window is too little, denoising effect is bad, and window is too large, can lose too many image detail again, causes the fuzzy of image.
The method of extreme value median filtering algorithm judgement noise is too simple, and extreme point is noise spot not necessarily, and noise spot also differs, and to establish a capital be extreme point.Therefore, in noise detection window, adopt bounding method judgement noise spot, even there is no noise in subrange, also can be mistaken for noise spot to some signaling points; Or in subrange, noise spot surpasses more than 2, and detected noise spot is not extreme point, easily causes noise spot detected undetected, can not well solve the fuzzy problem of image detail.
The weighted median filtering algorithm has weakened the ability of medium filtering elimination noise, simultaneously because the selection weights that before using, the very important person is have strengthened the difficulty of use.
Adaptive filter algorithm in actual applications, the maximum window size that the adaptive median filter algorithm be difficult to select is fit to selects too littlely, does not reach filter effect; Select too greatly, the algorithm calculated amount can increase significantly, and is consuming time oversize.And under certain condition, even increase window size, filter effect can not improve significantly yet again.
In follow-on median filtering algorithm, when carrying out walkaway, the point that will fall into interval [0, T], [T, 255] directly is defined as noise, but, in some image, [0, T], [T, 255] interval interior point might be also pixel, and the more flase drop that easily causes noise spot of such pixel; Due to image by noise pollution after, increase and decrease has in various degree mostly occured in the pixel value of signal, therefore, even between adjacent signaling point, some differences are larger, and it is less that the difference between noise spot and signaling point has, and during the noise secondary detection, Difference of Adjacent Pixels method threshold value is single, utilizes this method to come the detection noise point, so also easily causes the flase drop of noise spot.
In the process of medium filtering, the noise point value of processing and the contrast of the pixel value of original image exist the error of some.In more existing median filtering algorithms, noise spot or processed noise spot all be added in the algorithm implementation procedure, easily cause the larger error of calculation, so that keep the detail section effect not reach satisfied effect.
Summary of the invention
The purpose of this invention is to provide a kind of adaptive multistage weighted median filtering algorithm that is applied to digital picture, easily cause the flase drop of loss in detail and walkaway, undetected problem in order to solve existing median filtering algorithm.
For achieving the above object, the solution of the present invention is: a kind of adaptive multistage weighted median filtering algorithm that is applied to digital picture, and step is as follows:
(1) according to the characteristics of picture noise value, noise spot in detected image, if in image, the pixel value of tested point equals 0, in 3 * 3 windows centered by this tested point, if there is the pixel value of 3 points to equal 0 and have more than 5 the pixel value of point fall into interval [0,15] in, tested point is signaling point, otherwise is noise spot;
If the tested point pixel value is not equal to 0, given one [0,29] the noise reference value in scope, if the absolute difference of the pixel value of tested point and described noise reference value is less than 20 or greater than 220, and there is the pixel value of point more than 5 to fall in interval [0,35] or [230,255] in 3 * 3 windows centered by tested point, tested point is signaling point, otherwise is noise spot;
(2) with detected noise spot in step (1) as pending noise spot, the row labels of going forward side by side, make the filter window of 5 * 5 centered by described pending noise spot, and filter window is divided into level, vertical, 45 °, 135 ° four directions;
(3) calculate the intermediate value of described 5 * 5 filter windows gray-scale value on level, vertical, 45 °, 135 ° four directions, if on the intermediate value that calculates and respective direction arbitrarily the gray-scale value of the point crossed of mark equate, the window step length on respective direction is increased by 2, calculate again the gray-scale value intermediate value, when the gray-scale value of the point that any mark is crossed until the intermediate value that calculates and the party make progress is all unequal, the length of window on record respective direction this moment;
(4) calculate the corresponding weighting coefficient of each pixel and the weighted value except the point that mark is crossed on described four direction, then calculate respectively the weights output valve on described four direction;
(5) according to the weights output valve on the described four direction that calculates in step (4), the weights output valve on four direction and pending noise point value are got intermediate value as the final output valve of pending noise spot in image.
Detected noise spot in step (1) is labeled as 1 or 0, and the noise spot of processing still is labeled as 1 or 0.
The method of calculating weighting coefficient on four direction in step (4) is:
Wherein Z (m, n) is the gray-scale value that in 5 * 5 zones, on respective direction, (m, n) puts,
Med is the intermediate value of 5 * 5 windows gray-scale value on respective direction, and L is the length of window on respective direction, and the weighting coefficient of noise spot equals 0.
The method of calculating weighted value output valve on four direction in step (4) is:
K=1,2,3,4, wherein, d (m, n)=Z (m, n) r (m, n), d (m, n) they are the weighted value of the gray-scale value Z (m, n) of (m, n) point on certain direction, the capable n row of (m, n) expression m.
In step (5) in image the computing method of the final output valve f (m, n) of pending noise spot be: weights output valve and pending noise point value on level, vertical, 45 °, 135 ° directions are got intermediate value, i.e. f (m, n)=med[f
k(m, n), Z (m, n)], k=1,2,3,4.
The beneficial effect that the present invention reaches: (1) the present invention is subject to the characteristics of pixel value after noise pollution by analysis image, find out the difference of noise spot and signaling point, comes the detection noise point by the difference between the noise point value, has improved the correctness that noise spot detects;
(2) the present invention is in the process that filtering algorithm is realized, detection window is divided into level, vertical, 45 °, 135 ° four directions, calculates respectively the weights on each direction, then get intermediate value, better keep the details characteristics of fringe region, approached to a greater extent original signal point pixel value;
(3) due to noise spot with processed noise spot and can strengthen error in the picture noise processing procedure, therefore, the present invention is in the process that filtering algorithm is realized, noise spot and processed noise spot and do not participate in calculating, can reduce the error in the picture noise processing procedure, thereby better keep the detail section of image.
Description of drawings
Fig. 1 is overview flow chart of the present invention;
Fig. 2 is walkaway schematic flow sheet of the present invention;
Fig. 3 is noise processed schematic flow sheet of the present invention;
Fig. 4 is noise processed window classification schematic diagram of the present invention;
Fig. 5 is that noise processed window self-adaptation of the present invention is adjusted schematic diagram.
Embodiment
Below in conjunction with accompanying drawing, the present invention is described in further detail:
As Fig. 1, method of the present invention comprises two steps, first detect the noise spot in image, use the adaptive multistage median filtering algorithm to process to detected noise spot again, that is: A. is first according to the characteristics of picture noise value, adopt many threshold methods detected image noise spot, detected noise spot is labeled as 1, noise detecting method is basically to be distributed in zone [0 according to noise figure, 29] or [230,255] the obvious characteristics of difference in and between the noise point value, a given specific noise reference value adopts many threshold methods detection noise point;
B. to detected noise spot, adopt 5 * 5 filter windows, window center point is pending noise spot, uses adaptive multistage weighted median filtering algorithm to calculate weights output valve on four direction.In the filtering algorithm implementation procedure, adopt adaptive multistage weighted median filtering method, at first detection window is divided into level, vertical, 45 °, 135 ° four directions, calculate in the intermediate value process on each direction, be whether that the value that is labeled as 1 point is come the window size on certain direction of adaptive change according to the intermediate value that calculates, calculate respectively the weights output valve on four direction, at last four weights and pending noise spot are got intermediate value as final output valve.
A. as Fig. 2, the concrete steps of many threshold methods detected image noise spot are as follows:
If A1. the pixel value of tested point is 0, use and do 3 * 3 detection windows centered by this tested point, if 3 point values are arranged in detection window is 0 and have the point more than 5 to fall in interval [0,15], this 0 value point is signaling point, otherwise is noise spot;
A2. the tested point pixel value is not 0, in the present embodiment, given noise reference value 10, if image slices vegetarian refreshments f (x, y) satisfies with the difference of reference value | f (x, y)-10|＜20, and satisfy with image slices vegetarian refreshments f (x simultaneously, y) centered by, 3 * 3 windows at place have the point more than 5 all to fall in interval [0,35], and this point is signaling point; If | f (x, y)-10|＜20 and the pixel that falls in interval [0,35] in 3 * 3 windows at place centered by image slices vegetarian refreshments f (x, y) are less than 5 (not comprising 5), and this is noise spot;
If A3. image slices vegetarian refreshments f (x, y) satisfies with the difference of reference value | f (x, y)-10|＞220, and with image slices vegetarian refreshments f (x, y) centered by, 3 * 3 windows at place have the point more than 5 all to fall in interval [230,255], and this point is signaling point; If | f (x, y)-10|＞220 and the pixel that falls in interval [230,255] in 3 * 3 windows at place centered by image slices vegetarian refreshments f (x, y) are less than 5 (not comprising 5), and this is noise spot;
A4. detected noise spot is marked, and be labeled as 1.
B. as Fig. 3, the concrete steps of detected noise spot being carried out adaptive multistage weighted median filtering algorithm are as follows:
B1. use 5 * 5 windows centered by pending noise spot, and this window is divided into level, vertical, 45 °, 135 ° four directions, as shown in Figure 4;
B2. the intermediate value med of the gray-scale value on the calculated level direction, if the intermediate value that calculates equals the party and makes progress that any one is labeled as the gray-scale value of 1 point, the window step length on horizontal direction is increased by 2, calculate again the gray-scale value intermediate value, until the intermediate value that calculates be not equal to that this side up any one be labeled as the gray-scale value of 1 noise spot;
B3. according to formula
Calculate respectively on horizontal direction the corresponding weighting coefficient r of each pixel (m, n) except being labeled as 1 point, mark be that the weighting coefficient r (m, n) of 1 equals 0, wherein,
Z (m, n) is the gray-scale value of (m, n) point on horizontal directions in 5 * 5 zones, when calculating the value of sum, is labeled as 1 point and does not participate in calculating.Its corresponding weighting coefficient r of gray-scale value Z (m, n) (m, n) of every bit on horizontal direction in 5 * 5 zones is multiplied each other, be designated as d (m, n), i.e. d (m, n)=Z (m, n) r (m, n).Weights output f on the terminal level direction
1For: f
1=∑ d (m, n).
B4. repeating step B1, B2, B3, B4 calculate respectively the weights output f on vertical, 45 °, 135 ° three directions
2, f
3, f
4
B5. the output valve of final pending noise spot is f (m, n)=med[f
k(m, n), Z (m, n)], k=1,2,3,4, wherein Z (m, n) is pending noise spot, the noise spot of processing on respective direction still is labeled as 1.
The detailed process of determining gray-scale value intermediate value on horizontal direction in step B2 is as follows:
If B2.1. the intermediate value on horizontal direction equals the party and makes progress that any one is labeled as the gray-scale value of 1 point, the step-length of the window that the party is made progress increases by 2, and namely the window step length of this moment is 7, as shown in Figure 5;
If B2.2. the step-length on horizontal direction is 7 o'clock, the intermediate value that calculates still equals the party and makes progress that certain is labeled as the gray-scale value of 1 point, and the step-length with the window on horizontal direction increases by 2, and namely the window step length that makes progress of the party is 9;
B2.3. namely in 5 * 5 detection windows, certain is labeled as the gray-scale value of 1 point if the intermediate value that calculates is for the party makes progress, and the window step length that the party is made progress increases to 5+2n, n=1,2,3 ..., any one is labeled as the gray-scale value of 1 point until the intermediate value that calculates is not for the party makes progress.
In other embodiments, the noise reference value can arbitrarily be chosen in [0,29] scope, and can be labeled as 0 to detected noise spot.
Claims (5)
1. an adaptive multistage weighted median filtering algorithm that is applied to digital picture, is characterized in that, step is as follows:
(1) according to the characteristics of picture noise value, noise spot in detected image, if in image, the pixel value of tested point equals 0, in 3 * 3 windows centered by this tested point, if there is the pixel value of 3 points to equal 0 and have more than 5 the pixel value of point fall into interval [0,15] in, tested point is signaling point, otherwise is noise spot;
If the tested point pixel value is not equal to 0, given one [0,29] the noise reference value in scope, if the absolute value of the pixel value of tested point and described noise reference value difference value is less than 20 or greater than 220, and there is the pixel value of point more than 5 to fall in interval [0,35] or [230,255] in 3 * 3 windows centered by tested point, tested point is signaling point, otherwise is noise spot;
(2) with detected noise spot in step (1) as pending noise spot, the row labels of going forward side by side, make the filter window of 5 * 5 centered by described pending noise spot, and filter window is divided into level, vertical, 45 °, 135 ° four directions;
(3) calculate the intermediate value of described 5 * 5 filter windows gray-scale value on level, vertical, 45 °, 135 ° four directions, if on the intermediate value that calculates and respective direction arbitrarily the gray-scale value of the point crossed of mark equate, the window step length on respective direction is increased by 2, calculate again the gray-scale value intermediate value, when the gray-scale value of the point that any mark is crossed until the intermediate value that calculates and the party make progress is all unequal, the length of window on record respective direction this moment;
(4) calculate the corresponding weighting coefficient of each pixel and the weighted value except the point that mark is crossed on described four direction, then calculate respectively the weights output valve on described four direction;
(5) according to the weights output valve on the described four direction that calculates in step (4), the weights output valve on four direction and pending noise point value are got intermediate value as the final filtering output value of pending noise spot in image.
2. the adaptive multistage weighted median filtering algorithm that is applied to digital picture according to claim 1, is characterized in that, detected noise spot in step (1) is labeled as 1 or 0, and the noise spot of processing still is labeled as 1 or 0.
3. the adaptive multistage weighted median filtering algorithm that is applied to digital picture according to claim 1, is characterized in that, the method for calculating the weighting coefficient of each point on four direction in step (4) is:
Wherein Z (m, n) is the gray-scale value that in 5 * 5 zones, on respective direction, (m, n) puts,
Med is the intermediate value of 5 * 5 windows gray-scale value on respective direction, and L is the length of window on respective direction, and the weighting coefficient of noise spot equals 0.
4. the adaptive multistage weighted median filtering algorithm that is applied to digital picture according to claim 1, is characterized in that, the method for calculating weights output valve on four direction in step (4) is: f
k=∑ d (m, n), k=1,2,3,4, wherein, d (m, n)=Z (m, n) r (m, n), d (m, n) they are the weighted value of the gray-scale value Z (m, n) of (m, n) point on respective direction, the capable n row of (m, n) expression m.
5. the multistage weighted median filtering algorithm of white adaptation that is applied to digital picture according to claim 1, it is characterized in that, the final output valve f (m of pending noise spot in image in step (5), n) computing method are: weights output valve and pending noise point value on level, vertical, 45 °, 135 ° directions are got intermediate value, be f (m, n)=med[f
k(m, n), Z (m, n)], k=1,2,3,4.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201310097363.7A CN103150733B (en) | 2013-03-25 | 2013-03-25 | Self-adapting multi-stage weighted median filtering algorithm applied to digital images |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201310097363.7A CN103150733B (en) | 2013-03-25 | 2013-03-25 | Self-adapting multi-stage weighted median filtering algorithm applied to digital images |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN103150733A true CN103150733A (en) | 2013-06-12 |
| CN103150733B CN103150733B (en) | 2015-07-15 |
Family
ID=48548786
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201310097363.7A Active CN103150733B (en) | 2013-03-25 | 2013-03-25 | Self-adapting multi-stage weighted median filtering algorithm applied to digital images |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN103150733B (en) |
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104020136A (en) * | 2014-06-25 | 2014-09-03 | 中国科学院重庆绿色智能技术研究院 | Method and device for denoising near infrared spectrum by wavelet mid-value |
| CN104599250A (en) * | 2015-01-23 | 2015-05-06 | 河南理工大学 | Local mean value different value based self-adaptive pseudo-sphere filtering method |
| CN105243649A (en) * | 2015-11-09 | 2016-01-13 | 天津大学 | Image denoising method based on secondary noise point detection |
| CN105913384A (en) * | 2016-03-21 | 2016-08-31 | 温州大学 | Real-time weighted median filtering method based on bilateral grid |
| CN105931192A (en) * | 2016-03-21 | 2016-09-07 | 温州大学 | Image texture filtering method based on weighted median filtering |
| CN106097292A (en) * | 2016-08-04 | 2016-11-09 | 甘宗平 | Sequential SAR space-time neighborhood Gauss Weighted median filtering speckle is made an uproar suppression fast algorithm |
| CN106333679A (en) * | 2016-09-21 | 2017-01-18 | 广州视源电子科技股份有限公司 | Electroencephalogram preprocessing method and electroencephalogram preprocessing system in sleep state analysis |
| CN109086758A (en) * | 2018-08-21 | 2018-12-25 | 东北大学 | A kind of industrial process abnormal area labeling method based on Weighted median filtering |
| CN110111273A (en) * | 2019-04-25 | 2019-08-09 | 四川轻化工大学 | A kind of restorative procedure of image |
| CN111325694A (en) * | 2020-02-25 | 2020-06-23 | 深圳市景阳科技股份有限公司 | Image noise removing method and device |
| CN112785513A (en) * | 2020-08-25 | 2021-05-11 | 青岛经济技术开发区海尔热水器有限公司 | Self-adaptive median filtering method for filtering impulse noise |
-
2013
- 2013-03-25 CN CN201310097363.7A patent/CN103150733B/en active Active
Non-Patent Citations (7)
| Title |
|---|
| GONZALO R. ARCE ET AL: "Detail-Preserving Ranked-Order Based Filters for Image Processing", 《IEEE TRANSACTIONS ON ACOUSTICS, SPEECH, AND SIGNAL PROCESSING》 * |
| JI-PING SUN ET AL: "Adaptive Filtering Algorithm Using Long-Range Correlation for the Removal of Impulse Noise from Images", 《PROCEEDINGS OF THE 6TH WORLD CONGRESS ON INTELLIGENT CONTROL AND AUTOMATION》 * |
| 倪麟 等: "基于自适应加权中值滤波的二维Otsu图像分割算法", 《计算机应用研究》 * |
| 王明常 等: "动态方向加权二维多级中值滤波的图像处理", 《吉林大学学报(信息科学版)》 * |
| 王明常 等: "基于方向加权二维多级中值滤波的遥感图像处理方法", 《测绘技术装备》 * |
| 钱晶洁 等: "加权多级中值滤波", 《山东工业大学学报》 * |
| 陈初侠 等: "去除椒盐噪声的自适应开关加权均值滤波", 《计算机工程》 * |
Cited By (20)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104020136B (en) * | 2014-06-25 | 2016-08-31 | 中国科学院重庆绿色智能技术研究院 | Small echo intermediate value near infrared spectrum denoising method and device |
| CN104020136A (en) * | 2014-06-25 | 2014-09-03 | 中国科学院重庆绿色智能技术研究院 | Method and device for denoising near infrared spectrum by wavelet mid-value |
| CN104599250A (en) * | 2015-01-23 | 2015-05-06 | 河南理工大学 | Local mean value different value based self-adaptive pseudo-sphere filtering method |
| CN105243649A (en) * | 2015-11-09 | 2016-01-13 | 天津大学 | Image denoising method based on secondary noise point detection |
| CN105913384A (en) * | 2016-03-21 | 2016-08-31 | 温州大学 | Real-time weighted median filtering method based on bilateral grid |
| CN105931192A (en) * | 2016-03-21 | 2016-09-07 | 温州大学 | Image texture filtering method based on weighted median filtering |
| CN105913384B (en) * | 2016-03-21 | 2017-08-15 | 温州大学 | Real-time Weighted median filtering method based on bilateral grid |
| CN105931192B (en) * | 2016-03-21 | 2017-08-15 | 温州大学 | Image texture filtering method based on Weighted median filtering |
| CN106097292B (en) * | 2016-08-04 | 2019-12-24 | 甘宗平 | Space-time neighborhood Gaussian weighted median filtering method for remote sensing image processing |
| CN106097292A (en) * | 2016-08-04 | 2016-11-09 | 甘宗平 | Sequential SAR space-time neighborhood Gauss Weighted median filtering speckle is made an uproar suppression fast algorithm |
| CN106333679A (en) * | 2016-09-21 | 2017-01-18 | 广州视源电子科技股份有限公司 | Electroencephalogram preprocessing method and electroencephalogram preprocessing system in sleep state analysis |
| CN106333679B (en) * | 2016-09-21 | 2019-03-01 | 广州视源电子科技股份有限公司 | EEG signals preprocess method and system in sleep state analysis |
| CN109086758A (en) * | 2018-08-21 | 2018-12-25 | 东北大学 | A kind of industrial process abnormal area labeling method based on Weighted median filtering |
| CN109086758B (en) * | 2018-08-21 | 2021-04-13 | 东北大学 | Industrial process abnormal area marking method based on weighted median filtering |
| CN110111273A (en) * | 2019-04-25 | 2019-08-09 | 四川轻化工大学 | A kind of restorative procedure of image |
| CN110111273B (en) * | 2019-04-25 | 2021-02-12 | 四川轻化工大学 | Image restoration method |
| CN111325694A (en) * | 2020-02-25 | 2020-06-23 | 深圳市景阳科技股份有限公司 | Image noise removing method and device |
| CN111325694B (en) * | 2020-02-25 | 2024-02-13 | 深圳市景阳科技股份有限公司 | Image noise removing method and device |
| CN112785513A (en) * | 2020-08-25 | 2021-05-11 | 青岛经济技术开发区海尔热水器有限公司 | Self-adaptive median filtering method for filtering impulse noise |
| CN112785513B (en) * | 2020-08-25 | 2023-04-18 | 青岛经济技术开发区海尔热水器有限公司 | Self-adaptive median filtering method for filtering impulse noise |
Also Published As
| Publication number | Publication date |
|---|---|
| CN103150733B (en) | 2015-07-15 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN103150733B (en) | Self-adapting multi-stage weighted median filtering algorithm applied to digital images | |
| CN105069394B (en) | Quick Response Code weighted average gray level method coding/decoding method and system | |
| CN107038416B (en) | Pedestrian detection method based on binary image improved HOG characteristics | |
| CN107392919B (en) | Adaptive genetic algorithm-based gray threshold acquisition method and image segmentation method | |
| CN115249246B (en) | Optical glass surface defect detection method | |
| CN110188730B (en) | MTCNN-based face detection and alignment method | |
| CN103475838B (en) | A kind of interlace-removing method based on edge self-adaption | |
| CN103093443B (en) | Based on the image salt-pepper noise adaptive filter method of GA-BP neural network | |
| CN112884712A (en) | Method and related device for classifying defects of display panel | |
| US20150003743A1 (en) | Object detection device and object detection method | |
| CN112907460B (en) | Remote sensing image enhancement method | |
| CN115049671A (en) | Cloth surface defect detection method and system based on computer vision | |
| CN108961301A (en) | It is a kind of based on the unsupervised Chaetoceros image partition method classified pixel-by-pixel | |
| CN112633132A (en) | Bearing fault diagnosis method and system based on two-dimensional vibration image enhancement | |
| CN114663392A (en) | Knowledge distillation-based industrial image defect detection method | |
| TW201810115A (en) | Fingerprint image detecting device and method | |
| CN114092456A (en) | Cell fluorescence image distinguishing method and system | |
| CN114140416A (en) | Glass edge detection method and system based on machine vision | |
| CN110349119B (en) | Pavement disease detection method and device based on edge detection neural network | |
| CN117058137A (en) | Visual inspection method for production quality of laminated glass | |
| CN110766662B (en) | Forging surface crack detection method based on multi-scale and multi-layer feature learning | |
| CN108491796A (en) | A kind of time domain period point target detecting method | |
| CN115359258B (en) | Weak and small target detection method and system for component uncertainty measurement | |
| CN103942792A (en) | Impurity detecting method in medicine detection robot based on time domain features of sequence images | |
| CN113643216B (en) | Welding image enhancement method based on computer vision |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant |