+

CN106373098B - Random impulse noise removal method based on dissimilar pixel statistics - Google Patents

Random impulse noise removal method based on dissimilar pixel statistics Download PDF

Info

Publication number
CN106373098B
CN106373098B CN201610769889.9A CN201610769889A CN106373098B CN 106373098 B CN106373098 B CN 106373098B CN 201610769889 A CN201610769889 A CN 201610769889A CN 106373098 B CN106373098 B CN 106373098B
Authority
CN
China
Prior art keywords
pixel
noise
image
pixels
window
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.)
Expired - Fee Related
Application number
CN201610769889.9A
Other languages
Chinese (zh)
Other versions
CN106373098A (en
Inventor
史再峰
许泽昊
庞科
曹清洁
高阳
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Tianjin University
Original Assignee
Tianjin University
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Tianjin University filed Critical Tianjin University
Priority to CN201610769889.9A priority Critical patent/CN106373098B/en
Publication of CN106373098A publication Critical patent/CN106373098A/en
Application granted granted Critical
Publication of CN106373098B publication Critical patent/CN106373098B/en
Expired - Fee Related legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T5/00—Image enhancement or restoration
    • G06T5/70—Denoising; Smoothing
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/20—Special algorithmic details
    • G06T2207/20024—Filtering details
    • G06T2207/20032—Median filtering

Landscapes

  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Image Processing (AREA)
  • Image Analysis (AREA)

Abstract

The present invention relates to image processing fields, to propose a kind of new random impulsive noise point detecting method, and there is preferable noise measuring ability, and the random impulsive noise that can be effectively removed in image to the median filtering of noise image application enhancements according to testing result.Thus, the technical solution adopted by the present invention is that, based on the method for suppressing random impulsive noise of non-similar pixel statistics, step is: determining the substantially distribution situation of image border first, and the difference according to edge degree of strength determines the threshold value T slightly determined for non-similar pixel；The pixel in image around each pixel is handled simultaneously, determine whether child window center pixel is the non-similar pixel for being detected pixel, the non-similar pixel number NDP on last statistic mixed-state point periphery determines whether the test point is noise by being compared with non-similar pixel number threshold value NDPCT；Median filter process is used to noise spot.Present invention is mainly applied to image processings.

Description

Method for suppressing random impulsive noise based on non-similar pixel statistics
Technical field
The present invention relates to image processing fields, more particularly to when carrying out random impulsive noise removal processing to image, right Problem is determined and filtered out in the differentiation of noise spot.Concretely relate to the image denoising method using block sorting detection noise.
Background technique
The random impulsive noise of image is mainly derived from the processes such as image acquisition, transmission and analog-to-digital conversion, its main feature is that making an uproar Sound position and gray value are that random and with surrounding point is discontinuous, it appears that seem random scatter on the image bright or dark Spot.Noise in image has seriously affected the quality of image, is effectively removed random impulsive noise present in image and makes an uproar Sound, while the detailed structure of image can be retained again, for example edge and texture are a major issues of image procossing research.Linearly Filtering tends not to random impulse noise mitigation, it is therefore desirable to use nonlinear filtering.Median filtering is a kind of most common non- Linear filter method, can preferably inhibit the random impulsive noise in image, but due to its there is no distinguish image pixel and Noise pixel handles all pixels point in image, filtered image image border more easy to be lost and detailed information.Such as What accurately determines the position of random impulsive noise, i.e. noise measuring is the committed step that random impulsive noise removes.
Summary of the invention
In order to overcome the deficiencies of the prior art, the present invention is directed to propose a kind of new random impulsive noise point detecting method, and According to testing result to the median filtering of noise image application enhancements, there is preferable noise measuring ability, and can be effective Ground removes the random impulsive noise in image.For this purpose, the technical solution adopted by the present invention is that, based on non-similar pixel statistics with Machine impulsive noise minimizing technology, step is: determining the substantially distribution situation of image border first, and according to edge degree of strength Difference determines the threshold value T slightly determined for non-similar pixel；Simultaneously to the pixel in image around each pixel at Reason, to be detected pixel xi,jCentered on establish the exposure mask of one (2N+1) × (2N+1), extract 8 from this big window A child window, the division mode that child window divides is arranged counterclockwise, these child windows include detected pixel；It counts respectively Calculate each child window central pixel point and tested measuring point xi,jAbsolute pixel gray scale difference value dk, subscript k represents k-th of window, right In child window Ω1, central pixel point xi-1.j-1, dkIs defined as:
dk=| xi,j-xi-1.j-1|
If meeting dk>=T continues to determine, pixel mean differential index AD is calculated in the child window, if Meet dk≥ADkDetermine that the child window center pixel is the non-similar pixel for being detected pixel, last statistic mixed-state point periphery Non- similar pixel number NDP, by be compared with non-similar pixel number threshold value NDPCT determine the test point whether be to make an uproar Sound；Remaining child window and so on；Median filter process is used to noise spot.
Difference according to edge degree of strength determines that the threshold value T slightly determined for non-similar pixel is comprised the concrete steps that, respectively Calculate the Ω of arranged counterclockwise1, Ω2, Ω3, Ω4The intermediate value of four child window all pixels point gray values is simultaneously ranked up, sequence Result afterwards is denoted as m1≤m2≤m3≤m4, roughly determined in this exposure mask according to the numerical value of this four gray values and sequence Edge distribution；Enable Δ ME=m4-m1,ΔMI=m3-m2, Δ MEWith Δ MIThis four sub- window pixel gray scale entirety differences are provided Measurement, Δ MEWith Δ MIValue be divided into three kinds of situations, and accordingly three kinds of situations judge this center pixel surrounding edge point Cloth situation, wherein carrying out Δ MEWith Δ MIThreshold value T is taken when comparing to determiner；The first situation indicates to work as Δ MI≥TrWhen, judgement There are stronger details or grain distribution in this window；Second situation indicates Δ MI≤TrAnd Δ ME≥TrWhen, that is, judge this There are part edge distribution, i.e., weak edge in window；The third situation indicates to work as Δ ME≤TrWhen, judge this window be it is smooth, There is no image border；Different threshold value T is respectively adopted to determine non-similar pixel in the different edge of degree of strength.
Pixel mean differential index AD is the index for measuring peripheral image vegetarian refreshments and central point mean differential, each sub- window Mouth central point pixel x0Surrounding has m neighborhood territory pixel, and neighborhood territory pixel value is xiThen AD is defined as
The result of noise measuring is charged to and is instructed in matrix, it is identical as image size to instruct matrix, and only there are two value, " 1 " representative image corresponding position is noise pixel point, and " 0 " representative image corresponding position is normal pixel, is adopted to noise image Image can be denoised with median filtering, due to having indicated the position of normal pixel in instructing matrix, when filtering is selected It takes the normal pixel point on noise pixel periphery and acquires intermediate value substitution current noise pixel.
When using median filter process to noise image, according to the noise measuring instructed in matrix as a result, by noise picture The gray value of element is set as the intermediate value of all normal pixel point gray values in 3 × 3 neighborhood window of noise spot.
Carry out repeated detection using the method for iteration and restore image to obtain preferable noise filtering effect, and into When the multiple noise filtering of row, neighbour of the noise pixel of last missing inspection comprising more approximate normal gray values in restoring image The number of domain pixel, the non-similar pixel in neighborhood increases therewith, and non-similar pixel number compares threshold value NDPCT also will be with The number of iterations increases and increases.
The features of the present invention and beneficial effect are:
1. the invention proposes a kind of improvement median filtering calculations for being detected and being removed for common random impulsive noise Method determines rough comparison threshold value, this both reduced when carrying out noise measuring according to the edge of varying strength or details first The child window number that pixel mean differential index calculates, and improve the verification and measurement ratio of impulsive noise.
2. improved median filter method is combined closely with testing result, when filtering, which weeds out, has been detected as noise Pixel improves the denoising effect at median filtering blurred picture edge, protects the details of image.
Detailed description of the invention:
Fig. 1 noise measuring flow chart.
Child window divides schematic diagram when Fig. 2 noise measuring.
Fig. 3 edge detection texture estimation schematic diagram.
(a) weak edge (c) smooth region of strong edge (b).
Specific embodiment
The invention proposes a kind of new random impulsive noise point detecting methods, and answer according to testing result noise image With improved median filtering, this method has preferable noise measuring ability, and can be effectively removed random in image Impulsive noise.
Fig. 1 is noise measuring flow chart, and the present invention determines the substantially distribution situation of image border first, and strong according to edge The difference of weak degree determines the threshold value T slightly determined for non-similar pixel；Simultaneously to the picture in image around each pixel Element is handled, to be detected pixel xi,jCentered on establish the exposure mask of one (2N+1) × (2N+1), 5 × 5 exposure mask is preferred But be not limited to the size of exposure mask, 8 child windows extracted from this big window, child window division methods as shown in Fig. 2, this A little window includes detected pixel.Calculate separately each child window central pixel point and tested measuring point xi,jAbsolute picture Plain gray scale difference value dk, with child window Ω1(central pixel point xi-1.j-1) for, d1Is defined as:
d1=| xi,j-xi-1.j-1|,
If meeting dk>=T continues to determine, pixel mean differential index AD is calculated in the child window, if Meet dk≥ADk, that is, can determine that the child window center pixel is the non-similar pixel for being detected pixel, subscript k is represented k-th Window, the non-similar pixel number NDP on last statistic mixed-state point periphery, by being carried out with non-similar pixel number threshold value NDPCT Compare and determines whether the test point is noise.
When edge detection, Ω in Fig. 2 is calculated separately1, Ω2, Ω3, Ω4In four child window all pixels point gray values It is worth and is ranked up, the result after sequence is denoted as m1≤m2≤m3≤m4, can be thick according to the numerical value of this four gray values and sequence Slightly determine the edge distribution in this exposure mask.Enable Δ ME=m4-m1,ΔMI=m3-m2, Δ MEWith Δ MIThis four can be provided The measurement of child window pixel grey scale entirety difference, Δ MEWith Δ MIValue be roughly divided into such as three kinds of situations of figure, and judge accordingly The distribution situation of this center pixel surrounding edge, wherein carrying out Δ MEWith Δ MIThreshold value T is taken when comparing to determiner.In figure (a) situation indicates to work as Δ MI≥TrWhen, it can determine whether there is stronger details or grain distribution in this window.(b) situation indicates in figure ΔMI≤TrAnd Δ ME≥TrWhen, that is, it can determine whether there is part edge distribution in this window, i.e., weak edge；(c) situation indicates in figure As Δ ME≤TrWhen, can determine whether this window be it is smooth, almost without image border；It adopts respectively at the different edge of degree of strength Non- similar pixel is determined with different threshold value T.
Pixel mean differential index AD is the index for measuring peripheral image vegetarian refreshments and central point mean differential.Each sub- window Mouth central point pixel x0Surrounding has m neighborhood territory pixel (m=8 here), and neighborhood territory pixel value is xi(1≤i≤8) then AD is defined as
The result of noise measuring is charged to and is instructed in matrix, it is identical as image size to instruct matrix, and only there are two value, " 1 " representative image corresponding position is noise pixel point, and " 0 " representative image corresponding position is normal pixel, is adopted to noise image Image can be denoised with improved median filtering, due to having indicated the position of normal pixel in instructing matrix, be filtered The normal pixel point on noise pixel periphery is chosen when wave and acquires intermediate value substitution current noise pixel.
During removing random impulsive noise, since noise position and gray value have randomness, one-time detection is past It is past to obtain accurate testing result, the noise pixel comprising many missing inspections, so the method using iteration is repeatedly examined Preferable noise filtering effect can be obtained by surveying and restoring image.And when carrying out multiple noise filtering, last missing inspection is made an uproar Acoustic image element includes the neighborhood territory pixel of more approximate normal gray values, of the non-similar pixel in neighborhood in restoring image Number increases therewith, causes normal pixel point by the generation of erroneous detection such case, non-similar pixel number to reduce successive ignition Comparing threshold value NDPCT will also increase as the number of iterations increases.
The implementation steps of the invention is divided to detection noise and removal two step of noise to carry out.It is random to being added when to noise measuring The image of impulsive noise is detected, and pixel is used by edge and noise measuring centered on each of image pixel Be grayscale image that contrast is 255, picture size is 512 × 512, and Size of Neighborhood is chosen for 5 × 5, threshold when edge detection Value Tr40 are set as, and uses different comparison threshold value T, i.e. smooth region, weak edge, strong edge ratio for the edge of varying strength It is respectively T1, T2, T3 compared with threshold value, meeting the following conditions T1≤T2≤T3., selection [10,25,40] is used as different sides respectively here The comparison threshold value of edge.Non- similar pixel number compares threshold value NDPCT and increases as the number of iterations n increases, and the number of iterations is more, It has more normal pixels and is erroneously detected as noise pixel, the edge of blurred picture, the present invention is used uniformly iteration time twice Number, in noise detecting process twice, non-similar pixel number compares threshold value NDPCT and is set to 5,6 to improve impulsive noise Verification and measurement ratio.
It, as a result, will according to the noise measuring instructed in matrix when using improved median filter process to noise image The gray value of noise pixel is set as the intermediate value of all normal pixel point gray values in 3 × 3 neighborhood window of noise spot.
The implementation steps of the invention is divided to detection noise and removal two step of noise to carry out.It is random to being added when to noise measuring The image of impulsive noise is detected, and pixel is used by edge and noise measuring centered on each of image pixel Be grayscale image that contrast is 255, picture size is 512 × 512, and Size of Neighborhood is chosen for 5 × 5, threshold when edge detection Value Tr40 are set as, and uses different comparison threshold value T, i.e. smooth region, weak edge, strong edge ratio for the edge of varying strength It is respectively T1, T2, T3 compared with threshold value, meeting the following conditions T1≤T2≤T3., selection [10,25,40] is used as different sides respectively here The comparison threshold value of edge.Non- similar pixel number compares threshold value NDPCT and increases as the number of iterations n increases, and the number of iterations is more, It has more normal pixels and is erroneously detected as noise pixel, the edge of blurred picture, the present invention is used uniformly iteration time twice Number, in noise detecting process twice, non-similar pixel number compares threshold value NDPCT and is set to 5,6 to improve impulsive noise Verification and measurement ratio.
It, as a result, will according to the noise measuring instructed in matrix when using improved median filter process to noise image The gray value of noise pixel is set as the intermediate value of all normal pixel point gray values in 3 × 3 neighborhood window of noise spot.

Claims (4)

1.一种基于非相似像素统计的随机脉冲噪声去除方法，其特征是，步骤是：首先确定图像边缘的大致分布情况，并依据边缘强弱程度的不同确定用于非相似像素粗判定的阈值T；同时对图像中每一个像素点周围的像素进行处理，以被检测像素点xi,j为中心建立一个(2N+1)×(2N+1)的掩膜，从这个大窗口中提取出8个子窗口，子窗口划分的划分方式是逆时针排列，这些子窗口均包含被检测像素点；分别计算各个子窗口中心像素点与被检测点xi,j的绝对像素灰度差值dk，下标k代表第k个窗口，对于子窗口Ω1，其中心像素点xi-1.j-1，dk定义为：1. a random impulse noise removal method based on dissimilar pixel statistics, it is characterized in that, the step is: first determine the approximate distribution of the image edge, and determine the threshold for the rough judgment of dissimilar pixels according to the difference in the degree of edge strength T; At the same time, the pixels around each pixel in the image are processed, and a (2N+1)×(2N+1) mask is established with the detected pixel x i, j as the center, and extracted from this large window 8 sub-windows are created, and the sub-windows are divided in a counterclockwise arrangement. These sub-windows all contain the detected pixels; respectively calculate the absolute pixel grayscale difference d between the center pixel of each sub-window and the detected points x i, j k , the subscript k represents the kth window, and for the sub-window Ω 1 , its center pixel x i-1.j-1 , d k is defined as: dk＝|xi,j-xi-1.j-1|d k =|x i,j -x i-1.j-1 | 如果满足dk≥T，继续进行判定，在该子窗口中计算像素平均差异性指标AD，如果满足dk≥ADk即判定该子窗口中心像素为被检测像素点的非相似像素，最后统计检测点周边的非相似像素个数NDP，通过与非相似像素个数阈值NDPCT进行比较确定该检测点是否为噪声；其余子窗口以此类推；对噪声点运用中值滤波处理；If d k ≥ T is satisfied, continue to judge, and calculate the pixel average difference index AD in the sub-window. If d k ≥ AD k is satisfied, it is determined that the central pixel of the sub-window is a non-similar pixel of the detected pixel point, and finally statistics The number of non-similar pixels NDP around the detection point is compared with the threshold NDPCT of the number of non-similar pixels to determine whether the detection point is noise; the rest of the sub-windows are analogous; the noise point is processed by median filtering; 像素平均差异性指标AD是衡量周边像素点与中心点平均差异性的指标，各个子窗口中心点像素x0周围有m个邻域像素，邻域像素值为xi则AD定义为：The pixel average difference index AD is an index to measure the average difference between the surrounding pixels and the center point. There are m neighborhood pixels around the center point pixel x 0 of each sub-window, and the neighborhood pixel value is x i , then AD is defined as: 将噪声检测的结果记入指导矩阵中，指导矩阵与图像大小相同，且只有两个值，“1”代表图像相应位置为噪声像素点，“0”代表图像相应位置为正常像素点，对噪声图像采用中值滤波即可对图像进行去噪，由于在指导矩阵中已经标明正常像素的位置，滤波时选取噪声像素周边的正常像素点并求得中值替代当前噪声像素点。The result of noise detection is recorded in the guidance matrix. The guidance matrix is the same size as the image and has only two values. "1" means that the corresponding position of the image is a noise pixel, and "0" means that the corresponding position of the image is a normal pixel. The image can be denoised by median filtering. Since the position of normal pixels has been marked in the guidance matrix, normal pixels around the noise pixels are selected during filtering and the median value is obtained to replace the current noise pixels. 2.如权利要求1所述的基于非相似像素统计的随机脉冲噪声去除方法，其特征是，依据边缘强弱程度的不同确定用于非相似像素粗判定的阈值T具体步骤是，分别计算逆时针排列的Ω1，Ω2，Ω3，Ω4四个子窗口所有像素点灰度值的中值并进行排序，排序后的结果记为m1≤m2≤m3≤m4，根据这四个灰度值的数值和排序粗略地确定这个掩膜中的边缘分布；令ΔME＝m4-m1,ΔMI＝m3-m2，ΔME和ΔMI提供这四个子窗口像素灰度整体差异的度量，ΔME和ΔMI的取值分为三种情况，并据此三种情况判断出这个中心像素周围边缘的分布情况，其中在进行ΔME和ΔMI的比较判定时取阈值Tr；第一种情况表示当ΔMI≥Tr时，判断这个窗口内有较强的细节或纹理分布；第二种情况表示ΔMI≤Tr而ΔME≥Tr时，即判断这个窗口内有部分边缘分布，即弱边缘；第三种情况表示当ΔME≤Tr时，判断这个窗口是平滑的，没有图像边缘；强弱程度不同的边缘分别采用不同的阈值T来判定非相似像素。2. the random impulse noise removal method based on dissimilar pixel statistics as claimed in claim 1, it is characterized in that, according to the difference of edge strength, determine the threshold value T that is used for dissimilar pixel rough judgment The concrete step is, calculate the inverse respectively. The median of the gray values of all the pixel points in the four sub-windows Ω 1 , Ω 2 , Ω 3 , and Ω 4 arranged clockwise is sorted, and the sorted result is recorded as m 1 ≤m 2 ≤m 3 ≤m 4 , according to this The numerical value and ordering of the four grayscale values roughly determine the edge distribution in this mask; let ΔME = m 4 -m 1 , ΔM I = m 3 -m 2 , ΔME and ΔMI provide the four sub-window pixels The measurement of the overall difference in gray level, the values of ΔME and ΔMI are divided into three cases, and the distribution of the edges around the central pixel is judged according to the three cases. When comparing and judging ΔME and ΔMI Take the threshold value Tr; the first case indicates that when ΔMI ≥ Tr , it is judged that there is a strong distribution of details or textures in this window; the second case indicates that ΔMI ≤ Tr and ΔME ≥ Tr , namely It is judged that there are some edge distributions in this window, that is, weak edges; the third case means that when ΔME ≤ T r , it is judged that the window is smooth and there is no image edge; the edges with different strengths and weaknesses use different thresholds T respectively. Determine dissimilar pixels. 3.如权利要求1所述的基于非相似像素统计的随机脉冲噪声去除方法，其特征是，在对噪声图像运用中值滤波处理时，根据指导矩阵中的噪声检测结果，将噪声像素的灰度值设置为该噪声点3×3邻域窗口内所有正常像素点灰度值的中值。3. The method for removing random impulse noise based on dissimilar pixel statistics as claimed in claim 1, characterized in that, when applying median filtering to the noise image, according to the noise detection result in the guidance matrix, the grayscale of the noise pixel is removed. The degree value is set as the median of the gray values of all normal pixels in the 3×3 neighborhood window of the noise point. 4.如权利要求1所述的基于非相似像素统计的随机脉冲噪声去除方法，其特征是，采用迭代的方法进行多次检测并恢复图像可以取得较好的噪声滤除效果，而在进行多次噪声滤除时，上一次漏检的噪声像素在恢复图像中包含更多的近似正常灰度值的邻域像素，其邻域中的非相似像素的个数随之增加，非相似像素个数比较阈值NDPCT也要随着迭代次数增加而增加。4. The method for removing random impulse noise based on dissimilar pixel statistics as claimed in claim 1, characterized in that, adopting an iterative method to perform multiple detections and restore the image can achieve better noise filtering effect, while performing multiple detections and restoring images. When the secondary noise is filtered out, the noise pixels that were missed last time contain more neighbor pixels with approximate normal gray values in the restored image, and the number of dissimilar pixels in the neighborhood increases accordingly, and the number of dissimilar pixels increases. The number comparison threshold NDPCT also increases with the number of iterations.
CN201610769889.9A 2016-08-30 2016-08-30 Random impulse noise removal method based on dissimilar pixel statistics Expired - Fee Related CN106373098B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201610769889.9A CN106373098B (en) 2016-08-30 2016-08-30 Random impulse noise removal method based on dissimilar pixel statistics

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201610769889.9A CN106373098B (en) 2016-08-30 2016-08-30 Random impulse noise removal method based on dissimilar pixel statistics

Publications (2)

Publication Number Publication Date
CN106373098A CN106373098A (en) 2017-02-01
CN106373098B true CN106373098B (en) 2019-04-23

Family

ID=57901918

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201610769889.9A Expired - Fee Related CN106373098B (en) 2016-08-30 2016-08-30 Random impulse noise removal method based on dissimilar pixel statistics

Country Status (1)

Country Link
CN (1) CN106373098B (en)

Families Citing this family (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109035173B (en) * 2018-08-15 2022-05-17 深圳大学 Image filtering method, storage medium and terminal device
CN109598723B (en) * 2018-12-11 2021-09-07 讯飞智元信息科技有限公司 Image noise detection method and device
CN109920113B (en) * 2019-03-13 2020-08-18 安徽龙运智能科技有限公司 Control method of intelligent lock system and intelligent lock system
CN110706171B (en) * 2019-09-26 2024-04-26 中国电子科技集团公司第十一研究所 Image noise reduction method and device
CN113628118B (en) * 2020-05-06 2023-12-08 北京君正集成电路股份有限公司 Method for denoising and filtering in flat area
CN112446838B (en) * 2020-11-24 2024-07-12 海南大学 Image noise detection method and device based on local statistical information
CN113920068B (en) * 2021-09-23 2022-12-30 北京医准智能科技有限公司 Body part detection method and device based on artificial intelligence and electronic equipment
CN116071220B (en) * 2023-03-06 2023-06-20 浙江华感科技有限公司 Image window data processing method, device, equipment and medium
CN116681703B (en) * 2023-08-03 2023-10-10 杭州鸿世电器股份有限公司 Intelligent switch quality rapid detection method
CN117676038B (en) * 2024-01-30 2024-04-05 北京点聚信息技术有限公司 Electronic license data secure sharing method and system

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101087365A (en) * 2006-06-10 2007-12-12 中兴通讯股份有限公司 A method for filtering image mixed noise
CN105243649A (en) * 2015-11-09 2016-01-13 天津大学 Image denoising method based on secondary noise point detection
CN105787902A (en) * 2016-03-22 2016-07-20 天津大学 Image noise reduction method which utilizes partitioning ordering to detect noise

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101087365A (en) * 2006-06-10 2007-12-12 中兴通讯股份有限公司 A method for filtering image mixed noise
CN105243649A (en) * 2015-11-09 2016-01-13 天津大学 Image denoising method based on secondary noise point detection
CN105787902A (en) * 2016-03-22 2016-07-20 天津大学 Image noise reduction method which utilizes partitioning ordering to detect noise

Non-Patent Citations (4)

* Cited by examiner, † Cited by third party
Title
A Universal Noise Removal Algorithm with an Impulse Detector;Roman Garnett 等;《IEEE Transactions on Image Processing》;20051231;第1-7页
一种去除图像混合噪声的新方法;杨辉 等;《衡阳师范学院学报》;20101231;第31卷(第6期);第36-39页
一种有效的混合噪声滤波算法;王建勇 等;《信息技术》;20051231;第48-50页
数字图像混合噪声滤除算法研究;陈秉涛;《中国优秀硕士学位论文全文数据库 信息科技辑》;20121015;正文第5章

Also Published As

Publication number Publication date
CN106373098A (en) 2017-02-01

Similar Documents

Publication Publication Date Title
CN106373098B (en) Random impulse noise removal method based on dissimilar pixel statistics
CN109870461B (en) Electronic components quality detection system
CN105787902B (en) Utilize the image denoising method of block sorting detection noise
Jafar et al. Efficient improvements on the BDND filtering algorithm for the removal of high-density impulse noise
CN102156996B (en) Image edge detection method
CN103778611B (en) Utilize the switch weight vectors median filter method of rim detection
CN114549441B (en) Straw defect detection method based on image processing
CN109472788B (en) A method for detecting flaws on the surface of aircraft rivets
CN103400357B (en) A kind of method removing image salt-pepper noise
CN113436216B (en) Electrical equipment infrared image edge detection method based on Canny operator
CN104899842B (en) The adaptive extreme value median filter method of sequence for remote line-structured light image
CN105225244A (en) Based on the noise detection method that minimum local mean square deviation calculates
CN106918602A (en) A kind of detection method of surface flaw based on machine vision of robust
CN117522778A (en) Hollow brick flaw detection system
CN102118547A (en) Image weighted filtering method
CN112330633A (en) A fault image segmentation method based on self-adaptive band-pass filtering for damaged jumper tape
CN102830045A (en) Fabric spray rating objective evaluating method based on image processing
CN105469413B (en) It is a kind of based on normalization ring weighting without refer to smear restoration image synthesis method for evaluating quality
JP2014238789A (en) Image processing program, image processing method and image processor
Kumar et al. Adaptive edge discriminated median filter to remove impulse noise
Karthikeyan et al. Hybrid approach of efficient decision-based algorithm and fuzzy logic for the removal of high density salt and pepper noise in images
Guo et al. An adaptive soft-morphological-gradient-filter-for-edge-detection
Russo New method for measuring the detail preservation of noise removal techniques in digital images
Chudasama et al. Survey on Various Edge Detection Techniques on Noisy Images
Ahmed Image enhancement and noise removal by using new spatial filters

Legal Events

Date Code Title Description
C06 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
CF01 Termination of patent right due to non-payment of annual fee

Granted publication date: 20190423

Termination date: 20190830

CF01 Termination of patent right due to non-payment of annual fee
点击 这是indexloc提供的php浏览器服务，不要输入任何密码和下载