CN110211126A - Image partition method based on intuitionistic fuzzy C mean cluster - Google Patents
Image partition method based on intuitionistic fuzzy C mean cluster Download PDFInfo
- Publication number
- CN110211126A CN110211126A CN201910507275.7A CN201910507275A CN110211126A CN 110211126 A CN110211126 A CN 110211126A CN 201910507275 A CN201910507275 A CN 201910507275A CN 110211126 A CN110211126 A CN 110211126A
- Authority
- CN
- China
- Prior art keywords
- pixel
- value
- image
- intuitionistic fuzzy
- cluster centre
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
- G06F18/232—Non-hierarchical techniques
- G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
- G06F18/23213—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions with fixed number of clusters, e.g. K-means clustering
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10024—Color image
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Probability & Statistics with Applications (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a kind of image partition methods based on intuitionistic fuzzy C mean cluster.Mainly solve the problems, such as in image segmentation to noise-sensitive, be easily trapped into local optimum, clusters number needs to set, scheme is: using based on switch mean value constructing tactics to the intuitionistic Fuzzy Sets of noise robustnessIt carries out curve fitting to the grey value profile of image, screens all peak points of matched curve as cluster centre initial value range set, and count maximum cluster centre quantity;On the basis of, using the linear weighting function coefficient in location information and grayscale information the construction intuitionistic fuzzy objective function of pixel, obtain subordinated-degree matrix U;U is evaluated according to classification range index evaluation index, obtains optimal subordinated-degree matrix, and be used for error detection Policy Filtering mistake point pixel and correctly classified.Invention enhances the robustness to noise, can adaptively determine image clustering number, can be used for the pretreatment of image recognition and computer vision.
Description
Technical field
The invention belongs to digital image processing fields, and in particular to arrive a kind of image partition method, can be used for image recognition
With the pretreatment of computer vision.
Background technique
From eighties of last century the seventies and eighties, numerous scholars give more sustained attention image segmentation, and image Segmentation Technology has become
The basic technology of various fields, as long as concerning image Segmentation Technology, image point cannot be lacked the problem of extracting content in image
The quality and effect for cutting work will directly or indirectly influence subsequent Image Engineering.Existing image partition method type is numerous
It is more, the image partition method based on threshold value, the image partition method based on edge, the image segmentation based on region can be summarized as
Method, image partition method based on cluster etc., conventional images cutting techniques are with the image partition method based on threshold value and are based on
Based on the image partition method of cluster, the image segmentation based on threshold value is a kind of method divided using demographic information, usually
Suitable threshold value is extracted, target part interested in image is separated.But its to be only applicable to gray difference apparent
Image, in addition, the interference elements such as noise also cause difficulty to the Accurate Segmentation work of image.Up to the present it is not formed still
A kind of unified theoretical system is applied to different fields, and image segmentation remains a popular research field.
Image is divided into according to the similitude between pixel in image by multiple classes based on the image partition method of cluster
Not, good with segmentation effect, time-consuming short, the features such as method is stable, becomes apply universal one of method in recent years.Often
Cluster segmentation method has: K mean value, Fuzzy C-Means Clustering, hierarchical cluster and spectral clustering) etc., wherein Fuzzy C-means
Cluster is one of the method being most widely used.Fuzzy C-mean algorithm method proposes by Dunne earliest, referring to: Dunne is a kind of and iteration
The relevant fuzzy image processing method and its application of Self Organization Analysis algorithm are in the compact easily separated cluster cybernetics journal of detection
.1973,3(3):32-57.(J.C.Dunn.A Fuzzy Relative of the ISODATA Process and Its
Use in Detecting Compact Well-Separated Clusters[J].Journal of Cybernetics,
1973,3 (3): 32-57.), this method is not due to accounting for any spatial information in image, to the noise-sensitive in image,
So that segmentation noisy image cannot get ideal segmentation effect.Later many scholars have delivered to be changed in conjunction with image space neighborhood information
Into method.Ahmed et al. space neighborhood information item is added in the objective function of fuzzy C-mean algorithm, proposes FCM_S algorithm,
Although the algorithm improves the robustness to noise, but computation complexity is higher, in order to which the calculating for reducing FCM_S algorithm is complicated
Neighborhood information is introduced into the objective function of algorithm by degree, Chen Songcan and Zhang Daoqiang by way of mean filter and median filtering
In, FCM_S1 and FCM_S2 algorithm is proposed, referring to: a kind of stable combination space based on kernel function Chen Songcan, Zhang Daoqiang
Volumes 34,1907 of transactions of information fuzzy C mean value image segmentation algorithm American Electronic electrical engineers' association system control processing-
1916,2004.(S.Chen and D.Zhang,“Robust Image Segmentation Using FCM with
Spatial Constraints Based on New Kernel-induced Distance Measure,”IEEE
Trans.Syst,Man,Cybern,vol.34,pp1907-1916,2004.);FCM_S1 algorithm is to Gauss in both algorithms
Noise processed effect is preferable, to salt-pepper noise poor robustness, though two FCM_S2 algorithms are preferable to salt-pepper noise treatment effect, but
But to Gaussian noise poor robustness, thus both algorithms cannot accomplish the robust universality to multiple types noise.Cai Weiling
Et al. combine image spatial information and grayscale information construct linear weighted function and image, propose and quickly generate FCM algorithm, join
See: a kind of fuzzy C-mean algorithm of the fast robust of the introducing local message for image segmentation of Cai Weiling, Chen Songcan, Zhang Daoqiang is poly-
Volumes of 40,825-838,2007. of class algorithm pattern-recognition (W.Cai, S.Chen, and D.Zhang, " Fast and Robust
Fuzzy C-means Clustering Algorithms Incorporating Local Information for Image
Segmentation, " Pattern Recognit., vol.40, no.3, pp.825-838, Mar.2007.), this method is right
Gaussian noise has and also has preferable robustness while higher robustness to salt-pepper noise;But algorithm above does not all account for
The more ambiguities of data, Cha Erre et al. is it has further been discovered that can be considered the more moulds of data using intuitionistic Fuzzy Sets theory
Paste property, it is more accurate to the classification of data, the fuzzy clustering method based on intuitionistic fuzzy data is proposed, referring to: a kind of novelty
Intuitionistic fuzzy C means clustering algorithm and its in medical image apply application soft com-puting (T, Chaira. " A novel
intuitionistic fuzzy C means clustering algorithm and its application to
medical images."Appl.Soft Comput.11(2):1711-1717,2011.);Due to being based on intuitionistic fuzzy data
Fuzzy clustering method it is also more sensitive to noise, Wei Erma et al. further by local spatial information introduce intuitionistic fuzzy C it is equal
In value-based algorithm, referring to: Wei Erma, Agrawal, improvement of the Sha Lan for the combination local spatial information of brain image segmentation
Intuitionistic fuzzy C mean algorithm application soft com-puting .543-557,2016. (H.Verma, R.K.Agrawal, A.Sharan, " An
Improved Intuitionistic Fuzzy C-means Clustering Algorithm Incorporating
Local Information for Brain Image Segmentation,”Appl.Soft Comput.,543–557,
2016)。
Although above-mentioned improved method optimizes performance of the fuzzy clustering algorithm on noiseproof feature to a certain extent, still
It is very sensitive to the initial value of cluster centre in the presence of not strong enough to noise robustness, it is unable to adaptive analysis image clustering number etc.
It is insufficient.
Summary of the invention
It is a kind of based on intuitionistic fuzzy C mean cluster it is an object of the invention in view of the above shortcomings of the prior art, provide
Image partition method improve the accuracy of separation to enhance to the robustness of noise, reduce the sensitivity to cluster centre initial value
Property, adaptively determine image clustering number.
To achieve the above object, technology of the invention includes the following steps:
(1) image to be split is inputted, if image to be split is color image, is first converted into gray level image;
(2) maximum number of iterations T=100, outage threshold ε=10 initiation parameter: are set-5, Fuzzy Weighting Exponent m=2, neighbour
Domain window radius ω=3, primary iteration number t=1, clusters number default initial values Y=2;
(3) using the intuitionistic Fuzzy Sets based on switch mean value constructing tactics to noise robustnessWherein, 1≤i≤n, n table
Show sum of all pixels mesh;
(4) it is directed to the grey value profile of input picture, is carried out curve fitting using least square method, selects matched curve
Cluster centre value of the gray level corresponding to peak point as initialization, and the range of estimation cluster numbers is set as { 2 ..., Ymax,
Cluster centre initial value range set C isWherein, cYIndicate the initial of the Y cluster centre
Value, 1≤Y≤Ymax, YmaxIndicate the maximum value of clusters number, 2≤Ymax≤ n, T indicate transposition;
(5) linear weighting function coefficient H is iterated to calculateir, Intuitionistic Fuzzy DistancesWith intuitionistic fuzzy objective function
J:
(5a) constructs linear weighting function coefficient H using the location information and grayscale information of pixelir:
Wherein, (ai, bi) and (ar, br) respectively indicate the spatial value of center pixel i and neighborhood territory pixel r, xiAnd xrPoint
Not Biao Shi center pixel i and neighborhood territory pixel r gray value, ukiIndicate xiFor k-th of cluster centre ckDegree of membership, 1≤k≤
Y, 1≤r≤ω2;
(5b) calculates pixel x under intuitionistic Fuzzy SetsiTo cluster centre ckIntuitionistic Fuzzy DistancesWherein,WithRespectively indicate pixel xiWith cluster centre ckIntuitionistic Fuzzy Sets;
The linear weighting function coefficient H that (5c) is obtained in conjunction with (5a)irThe Intuitionistic Fuzzy Distances obtained with (5b)
Intuitionistic fuzzy objective function J is calculated, and solves the subordinating degree function u in intuitionistic fuzzy objective function JkiWith cluster centre ckIt is straight
Feel fuzzy set
(6) judge iteration stopping condition | | C(t+1)-C(t)| | whether < ε or operation the number of iterations t > T are true:
If invalid, t=t+1 is enabled, is returned (5);
If set up, exports subordinated-degree matrix U and judge Y < YmaxIt is whether true, if so, Y=Y+1 is then enabled, cluster is taken
The preceding Y numerical value of center initial value range set C, enables the initial value set of cluster centre bePrimary iteration number t=1 is enabled, is returned (5), if Y < YmaxIt is invalid, it executes (7),
Wherein,For 0 cluster centre numerical value set,Indicate the Q cluster centre iteration
The t times intuitionistic Fuzzy Sets,Indicate the initial value of the P cluster centre, 1≤Q≤Y, 1≤P≤Y, U={ uki};
(7) (Y is inputted from (6)max- 1) a subordinated-degree matrix U, and these are evaluated using classification range index evaluation index GD
Subordinated-degree matrix;If the maximum subordinated-degree matrix of GD numerical value is Umax, by UmaxCorresponding clusters number YfinAs input picture
Final clusters number, wherein 1≤Yfin≤Ymax;
(8) to the subordinated-degree matrix U of (7) outputmaxPixel classifications are done, and carry out error detection classification, export segmented image:
(8a) is according to maximum membership grade principle and final clusters number Yfin, to the subordinated-degree matrix U of (7) outputmaxPicture
Element is classified, output label matrix UL, wherein ULInside contain only categorization values { 1,2 ..., YW..., Yfin, 1≤YW≤Yfin;
(8b) sets the label class label of center pixel i as Li, multiply in 5 neighborhoods the 5 of center pixel i, judge label classification
Value LiPixel quantity nLWhether n is metL≤25/(Yfin+ 1) condition:
If it is satisfied, then determining that center pixel i divides pixel for mistake, and mistake is divided to the label class label L of pixeliIt is modified to neighbour
The most label class label of label class label quantity of the same race in domain executes (8c);
If conditions are not met, then determining that center pixel i is correct distribution pixel, execute (8c);
(8c) judges label matrix ULIn pixel whether all traverse, if so, exporting final label matrix UF, execute
(8d) continues to traverse if it is not, returning to (8b);
(8d) is to final label matrix UFMiddle label class label is YWPixel assign different corresponding gray value xW, and will
The matrix for assigning different gray values is exported as segmented image.
Compared with prior art, the invention has the following beneficial technical effects:
First, the present invention is used based on switch mean value constructing tactics to the intuitionistic Fuzzy Sets of noise robustness, utilizes pixel
Location information and grayscale information construction linear weighting function coefficient are simultaneously introduced into intuitionistic fuzzy objective function, are solved cluster and are calculated
The problem of method is to noise-sensitive.
Second, the present invention is carried out curve fitting using grey value profile of the least square method to image, and selects fitting bent
Gray level corresponding to all peak points of line is commented as cluster centre initial value range according to classification range index evaluation index
Valence subordinated-degree matrix selects the clusters number for being suitble to input picture, realizes adaptively determining for image clustering number, and overcome
The shortcomings that traditional clustering algorithm is easily trapped into local optimum.
Third, the present invention divides pixel using error detection Policy Filtering mistake, and divides element correctly to be distributed mistake, improves point
Accuracy is cut, so that segmentation effect is even more ideal.
Detailed description of the invention
Fig. 1 is implementation flow chart of the invention;
Fig. 2 is to carry out emulation segmentation to the #3096 image in Berkeley image data base with the present invention and existing method
Comparative result figure;
Fig. 3 is to carry out emulation point to the #238011 image in Berkeley image data base with the present invention and existing method
The comparative result figure cut.
Specific embodiment
The embodiment of the present invention and effect are further described below in conjunction with attached drawing:
Referring to Fig. 1, steps are as follows for realization of the invention:
Step 1: inputting image to be split.
Image to be split is inputted, if image to be split is color image, is first converted into gray level image.
Step 2: setting initial parameter value.
If maximum number of iterations T=100, outage threshold ε=10-5, Fuzzy Weighting Exponent m=2, neighborhood window radius ω=3,
Primary iteration number t=1, clusters number default initial values Y=2.
Step 3: constructing the intuitionistic Fuzzy Sets to noise robustness
The method of prior art configuration intuitionistic Fuzzy Sets includes IFCM algorithm and IIFCM algorithm, is calculated using Yager
Son constructs intuitionistic Fuzzy Sets.
This example use based on switch mean value strategy amendment input picture grey scale pixel value, and by Yager operator come
Construct the intuitionistic Fuzzy Sets to noise robustness, wherein switch mean value strategy is according to the pixel grey scale Distribution value in pixel i neighborhood
Situation does gray value amendment, is accomplished by
3.1) first determine whether the grey scale pixel value multiplied in 3 neighborhoods the 3 of center pixel i is not all 0 or 255:
If set up, then it represents that there are also other gray values in addition to 0 or 255 for the grey scale pixel value in neighborhood, first by neighborhood
The pixel that interior gray value is 0 and 255 is taken out, then takes the gray average of rest of pixels in neighborhood, and be set as the amendment of center pixel
Numerical value, then execute 3.2);
If invalid, the gray value of center pixel is not modified, is directly executed 3.2);
3.2) judge whether the pixel in input picture all traverses: if so, the revised image of output gray level value, holds
Row 3.3), 3.1) continue to traverse if it is not, returning;
3.3) 3.2) grey scale pixel value of the amendment image of output is calculated as intuitionistic Fuzzy Sets
3.3.1 μ (x) is seti) indicate grey scale pixel value xiDegree of membership, v (xi) and π (xi) it is x respectivelyiUse Yager operator
The non-affiliated degree and hesitation degree of construction, μ (xi)、v(xi) and π (xi) calculation formula difference it is as follows:
π(xi)=1- μ (xi)-v(xi),
Wherein, 0≤μ (xi)≤1,0≤v (xi)≤1,0≤μ (xi)+v(xi)≤1, α=0.8 indicates Yager operator system
Number;
3.3.2) according to 3.3.1) μ (x that is calculatedi)、v(xi) and π (xi) construction intuitionistic Fuzzy Sets
Step 4: for the grey value profile of input picture, being carried out curve fitting using least square method, selection fitting is bent
Cluster centre value of the gray level corresponding to the peak point of line as initialization estimates that maximum clusters number and cluster centre are initial
It is worth range.
The existing technology for choosing cluster centre initial value has the peak point for directly choosing image grey level histogram as initial
Cluster centre.
This example is first carried out curve fitting to image grayscale Distribution value using least square method, then bent in smooth fitting
Line options peak point, is accomplished by
4.1) gray level image is set as 256 gray levels, grey level range is [0,255], and draws the ash of gray level image
Spend histogram;
4.2) number of pixels corresponding to 256 gray levels on grey level histogram is counted, grey value profile frequency f is set as
(l), curve matching and using seven rank least square methods is done to f (l), gray value data point is fitted to gray value curve, wherein
0≤l≤255;
4.3) the first derivative f ' (l) of f (l) is sought, and compares the first derivative f ' (l- of gray level before and after f ' (l) and f (l)
And f ' (l+1) 1);
4.4) judge gray level l whether under conditions of l=0 l=255:
It, will if set up if under conditions of l=0 l=255 and l=0, then judging whether f ' (l) < 0 is true
Gray level l=0 is exported as peak point, then is executed 4.5);
If under conditions of l=0 l=255 and l=255, then judging whether f ' (l) > 0 is true, if set up,
It exports, then is executed 4.5) using gray level l=255 as peak point;
If not indicating under conditions of 0 255 < < l under conditions of l=0 l=255, then judging f ' (l)
Whether=0 and f ' (l-1) > 0 and f ' (l+1) < 0 are true, if set up, export gray level l as peak point, then execute
4.5);
4.5) judge whether 256 gray levels all traverse, if so, 4.6) output peak point, executes, if it is not,
4.4) return continues to traverse;
4.6) it sorts to the peak point of 4.5) output according to the sequence of intensity profile frequency values from big to small, counts peak point
Number be YmaxIf cluster numbers range is 2 to YmaxIt is a, and the corresponding gray level l of peak point sorted from large to small is counted, it obtains
It is to gray scale value rangeIt is cluster centre initial value that gray level, which is corresponded to numerical statistic, i.e.,If cluster centre initial value range set C isWherein:
lYIndicate that in peak point intensity profile frequency values, the Y peak point intensity profile is frequently according in descending order
Gray level corresponding to rate value, cYIndicate the initial value of the Y cluster centre, 1≤Y≤Ymax, YmaxIndicate clusters number most
Big value, 2≤Ymax≤ n, T indicate transposition.
Step 5: iterative calculation linear weighting function coefficient Hir, Intuitionistic Fuzzy DistancesWith intuitionistic fuzzy target letter
Number J.
The local spatial information of image is introduced into its objective function by existing FCM_S1 algorithm and FCM_S2 algorithm,
In, what FCM_S1 algorithm considered is the mean filter feature of image, and what FCM_S2 algorithm considered is the median filtering feature of image.
This example constructs linear weighting function coefficient H first with the location information and grayscale information of pixelir, then will include
The H of image local spatial informationirIt is introduced into intuitionistic fuzzy objective function J, is accomplished by
5.1) linear weighting function coefficient H is constructed using the location information of pixel and grayscale informationir:
Wherein, (ai,bi) and (ar,br) respectively indicate the spatial value of center pixel i and neighborhood territory pixel r, xiAnd xrPoint
Not Biao Shi center pixel i and neighborhood territory pixel r gray value, ukiIndicate xiFor k-th of cluster centre ckDegree of membership, 1≤k≤
Y, 1≤r≤ω2;
5.2) grey scale pixel value x under intuitionistic Fuzzy Sets is calculatediTo cluster centre ckIntuitionistic Fuzzy Distances
5.2.1 μ (c) is setk) indicate cluster centre ckDegree of membership, v (ck) and π (ck) respectively indicate cluster centre ckIt uses
The non-affiliated degree and hesitation degree of Yager operator construction, μ (ck)、v(ck) and π (ck) calculation formula difference it is as follows:
π(ck)=1- μ (ck)-v(ck),
Wherein, 0≤μ (ck)≤1,0≤v (ck)≤1,0≤μ (ck)+v(ck)≤1;
5.2.2 Intuitionistic Fuzzy Distances) are calculated according to the result that (5.2.1) is obtained
Wherein, μ (xi)、v(xi) and π (xi) respectively indicate grey scale pixel value xiDegree of membership, non-affiliated degree and hesitation degree;
5.3) the linear weighting function coefficient H obtained in conjunction with (5.1)irThe Intuitionistic Fuzzy Distances obtained with (5.2)Calculate intuitionistic fuzzy objective function J:
Wherein, ukiIt is pixel xiTo cluster centre ckSubordinating degree function, m indicate FUZZY WEIGHTED index;
5.4) intuitionistic fuzzy objective function J is solved using lagrange's method of multipliers, and obtains subordinating degree function ukiAnd cluster
Center ckIntuitionistic Fuzzy Sets
5.4.1 Lagrange multiplier λ) is introduced in intuitionistic fuzzy objective function Ji, constructing has Lagrange multiplier
Objective function L:
5.4.2 the objective function L with Lagrange multiplier) is calculated separately for subordinating degree function uki, cluster centre ck
Intuitionistic Fuzzy SetsWith Lagrange multiplier λiPartial derivativeWith
5.4.3) in partial derivativeUnder conditions of, by acquiring subordinating degree function u as followskiWith cluster
Center ckIntuitionistic Fuzzy Sets
Wherein, k=1,2 ..., Y, i=1,2 ..., n, r=1,2 ..., ω2。
Step 6: judging whether to meet iteration stopping condition.
If meeting iteration stopping condition | | C(t+1)-C(t)| | < ε or operation the number of iterations t > T then exports degree of membership square
Battle array U, and execute 7), whereinFor cluster centre numerical value set,Indicate Q
The intuitionistic Fuzzy Sets that cluster centre iteration is the t times;
If being unsatisfactory for iteration stopping condition, t=t+1 is enabled, is returned 5.1).
Step 7: judging whether current clusters number Y meets condition.
If meeting current clusters number Y is less than maximum clusters number Ymax, i.e. Y < Ymax, then Y=Y+1 is enabled, is then taken poly-
The preceding Y numerical value of class center initial value range set C, is output to the initial value set of cluster centre, and enable cluster centre initial value
Set C(1)ForPrimary iteration number t=1 is enabled, is returned 5.1), whereinIndicate P
The initial value of a cluster centre, 1≤P≤Y;
If being unsatisfactory for Y < Ymax, then (8) are executed.
Step 8: evaluating subordinated-degree matrix using classification range index evaluation index GD.
The existing evaluation index to subordinated-degree matrix has division factor VpcWith partition entropy Vpe, can reflect degree of membership square
The fog-level of battle array, wherein VpcNumerical value is bigger, and the ambiguity of subordinated-degree matrix is smaller, VpeNumerical value is smaller, and pixel classifications are more quasi-
Really.
This example evaluates subordinated-degree matrix using classification range index evaluation index GD, i.e., inputs (Y from 7) middlemax- 1) a
Subordinated-degree matrix U evaluates these subordinated-degree matrix using classification range index evaluation index GD, and GD numerical value is maximum is subordinate to for output
Matrix is spent, and sets the maximum subordinated-degree matrix of GD numerical value as Umax, by UmaxCorresponding clusters number YfinMost as input picture
Whole clusters number, in which: classification range index evaluation index GD formula is as follows:
Wherein, ui1It indicates that ith pixel is maximum and is subordinate to angle value, ui2Indicate ith pixel it is second largest be subordinate to angle value, 1
≤Yfin≤Ymax。
Step 9: to subordinated-degree matrix UmaxCarry out pixel classifications.
According to maximum membership grade principle and final clusters number Yfin, to the subordinated-degree matrix U of 8) outputmaxPixel carry out
Classification, output label matrix UL, wherein maximum membership grade principle is according to Vk=argk{max(uki) pixel i is divided into person in servitude
The maximum classification V of category degreek, ULInside contain only categorization values { 1,2 ..., YW..., Yfin, 1≤YW≤Yfin。
Step 10: to label matrix ULCarry out error detection classification.
10.1) the label matrix U of 9) output is setLCenter pixel i label class label be Li, 5 in center pixel i multiply 5
In neighborhood, label class label L is judgediPixel quantity nLWhether n is metL≤25/(Yfin+ 1) condition:
If it is satisfied, then determining that center pixel i divides pixel for mistake, and mistake is divided to the label class label L of pixeljIt is modified to neighbour
10.2) the most label class label of label class label quantity of the same race in domain executes;
If conditions are not met, then determining that center pixel i is correct distribution pixel, execute 10.2);
10.2) judge label matrix ULIn pixel whether all traverse, if so, exporting final label matrix UF, execute
11), 10.1) continue to traverse if it is not, returning.
Step 11: to final label matrix UFAssignment simultaneously exports image.
To the final label matrix U of 10.2) outputFMiddle label class label is YWPixel assign different corresponding gray values
xW, and the matrix for assigning different gray values is exported as segmented image, final label matrix UFInside contain only categorization values 1,
2 ..., YW..., Yfin, assignment formula is as follows:
xW=[255/YW],
Wherein, [] indicates to take internal integer.
Below in conjunction with emulation experiment, technical effect of the invention is described further;
1. simulated conditions:
Emulation experiment is in computer Inter (R) Core (TM) i5-3210M 2.50GHZ CPU, 8G memory, MATLAB
It is carried out under R2014a software environment.
2. emulation content:
The image that the number in Berkeley image data base is 3096 is chosen in emulation 1, with the present invention and the existing side FCM
Method, FCM_S1 method, FCM_S2 method, FGFCM method, IFCM method and IIFCM method are respectively split it, as a result such as
Shown in Fig. 2, in which:
2 (a) be the original image of 3096 images;
2 (b) be the Standard Segmentation figure of 3096 images;
2 (c) be the Gauss noisy image of 3096 images, noise intensity 0.006;
2 (d) be the segmentation result with existing FCM method to the Gauss noisy image of 3096 images;
2 (e) be the segmentation result with existing FCM_S1 method to the Gauss noisy image of 3096 images;
2 (f) be the segmentation result with existing FCM_S2 method to the Gauss noisy image of 3096 images;
2 (g) be the segmentation result with existing FGFCM method to the Gauss noisy image of 3096 images;
2 (h) be the segmentation result with existing IFCM method to the Gauss noisy image of 3096 images;
2 (i) be the segmentation result with existing IIFCM method to the Gauss noisy image of 3096 images;
2 (j) be the segmentation result with the present invention to the Gauss noisy image of 3096 images;
2 (k) be the spiced salt noisy image of 3096 images, noise intensity 0.01;
2 (l) be the segmentation result with existing FCM method to the spiced salt noisy image of 3096 images;
2 (m) be the segmentation result with existing FCM_S1 method to the spiced salt noisy image of 3096 images;
2 (n) be the segmentation result with existing FCM_S2 method to the spiced salt noisy image of 3096 images;
2 (o) be the segmentation result with existing FGFCM method to the spiced salt noisy image of 3096 images;
2 (p) be the segmentation result with existing IFCM method to the spiced salt noisy image of 3096 images;
2 (q) be the segmentation result with existing IIFCM method to the spiced salt noisy image of 3096 images;
2 (r) be the segmentation result with the present invention to the spiced salt noisy image of 3096 images;
Figure it is seen that the present invention can separate background and target on the basis of inhibiting noise, image is poly-
Class number can be determined adaptively, insensitive to the initial value of cluster centre, so segmentation effect of the present invention to noisy image
Better than existing FCM method, FCM_S1 method, FCM_S2 method, FGFCM method, IFCM method and IIFCM method.
The image that the number in Berkeley image data base is 238011 is chosen in emulation 2, with the present invention and existing FCM
Method, FCM_S1 method, FCM_S2 method, FGFCM method, IFCM method and IIFCM method are respectively split it, as a result
As shown in Figure 3, in which:
3 (a) be the original image of 238011 images;
3 (b) be the Standard Segmentation figure of 238011 images;
3 (c) be the Gauss noisy image of 238011 images, noise intensity 0.006;
3 (d) be the segmentation result with existing FCM method to the Gauss noisy image of 238011 images;
3 (e) be the segmentation result with existing FCM_S1 method to the Gauss noisy image of 238011 images;
3 (f) be the segmentation result with existing FCM_S2 method to the Gauss noisy image of 238011 images;
3 (g) be the segmentation result with existing FGFCM method to the Gauss noisy image of 238011 images;
3 (h) be the segmentation result with existing IFCM method to the Gauss noisy image of 238011 images;
3 (i) be the segmentation result with existing IIFCM method to the Gauss noisy image of 238011 images;
3 (j) be the segmentation result with the present invention to the Gauss noisy image of 238011 images;
3 (k) be the spiced salt noisy image of 238011 images, noise intensity 0.01;
3 (l) be the segmentation result with existing FCM method to the spiced salt noisy image of 238011 images;
3 (m) be the segmentation result with existing FCM_S1 method to the spiced salt noisy image of 238011 images;
3 (n) be the segmentation result with existing FCM_S2 method to the spiced salt noisy image of 238011 images;
3 (o) be the segmentation result with existing FGFCM method to the spiced salt noisy image of 238011 images;
3 (p) be the segmentation result with existing IFCM method to the spiced salt noisy image of 238011 images;
3 (q) be the segmentation result with existing IIFCM method to the spiced salt noisy image of 238011 images;
3 (r) be the segmentation result with the present invention to the spiced salt noisy image of 238011 images;
From figure 3, it can be seen that the present invention can separate background and target on the basis of inhibiting noise, image is poly-
Class number can be determined adaptively, insensitive to the initial value of cluster centre, be better than to the segmentation effect of noisy image existing
FCM method, FCM_S1 method, FCM_S2 method, FGFCM method, IFCM method and IIFCM method.
Claims (9)
1. the image partition method based on intuitionistic fuzzy C mean cluster, which is characterized in that include the following:
(1) image to be split is inputted, if image to be split is color image, is first converted into gray level image;
(2) maximum number of iterations T=100, outage threshold ε=10 initiation parameter: are set-5, Fuzzy Weighting Exponent m=2, neighborhood window
Radius ω=3, primary iteration number t=1, clusters number default initial values Y=2;
(3) using the intuitionistic Fuzzy Sets based on switch mean value constructing tactics to noise robustnessWherein, 1≤i≤n, n indicate picture
Plain total number;
(4) it is directed to the grey value profile of input picture, is carried out curve fitting using least square method, selects the peak value of matched curve
Cluster centre value of the corresponding gray level of point as initialization, and the range of estimation cluster numbers is set as { 2 ..., Ymax, cluster
Center initial value range set C isWherein, cYIndicate the initial value of the Y cluster centre,
1≤Y≤Ymax, YmaxIndicate the maximum value of clusters number, 2≤Ymax≤ n, T indicate transposition;
(5) linear weighting function coefficient H is iterated to calculateir, Intuitionistic Fuzzy DistancesWith intuitionistic fuzzy objective function J:
(5a) constructs linear weighting function coefficient H using the location information and grayscale information of pixelir:
Wherein, (ai,bi) and (ar,br) respectively indicate the spatial value of center pixel i and neighborhood territory pixel r, xiAnd xrTable respectively
Show the gray value of center pixel i and neighborhood territory pixel r, ukiIndicate xiFor k-th of cluster centre ckDegree of membership, 1≤k≤Y, 1
≤r≤ω2;
(5b) calculates pixel x under intuitionistic Fuzzy SetsiTo the Intuitionistic Fuzzy Distances of cluster centre ckWherein,With
Respectively indicate pixel xiWith the intuitionistic Fuzzy Sets of cluster centre ck;
The linear weighting function coefficient H that (5c) is obtained in conjunction with (5a)irThe Intuitionistic Fuzzy Distances obtained with (5b)It calculates
Intuitionistic fuzzy objective function J, and solve the subordinating degree function u in intuitionistic fuzzy objective function JkiWith cluster centre ckIntuition mould
Paste collection
(6) judge iteration stopping condition | | C(t+1)-C(t)| | whether < ε or operation the number of iterations t > T are true:
If invalid, t=t+1 is enabled, is returned (5);
If set up, exports subordinated-degree matrix U and judge Y < YmaxIt is whether true, if so, Y=Y+1 is then enabled, cluster centre is taken
The preceding Y numerical value of initial value range set C, enables the initial value set of cluster centre beIt enables
Primary iteration number t=1 is returned (5), if Y < YmaxIt is invalid, it executes (7), whereinFor 0 cluster centre numerical value set,Indicate the Q cluster centre iteration t
Secondary intuitionistic Fuzzy Sets,Indicate the initial value of the P cluster centre, 1≤Q≤Y, 1≤P≤Y, U={ uki};
(7) (Y is inputted from (6)max- 1) a subordinated-degree matrix U, and evaluate these using classification range index evaluation index GD and be subordinate to
Spend matrix;If the maximum subordinated-degree matrix of GD numerical value is Umax, by UmaxCorresponding clusters number YfinAs the final of input picture
Clusters number, wherein 1≤Yfin≤Ymax;
(8) to the subordinated-degree matrix U of (7) outputmaxPixel classifications are done, and carry out error detection classification, export segmented image:
(8a) is according to maximum membership grade principle and final clusters number Yfin, to the subordinated-degree matrix U of (7) outputmaxPixel into
Row classification, output label matrix UL, wherein ULInside contain only categorization values 1,2 ..., YW..., Yfin, 1≤YW≤Yfin;
(8b) sets the label class label of center pixel i as Li, multiply in 5 neighborhoods the 5 of center pixel i, judge label class label Li's
Pixel quantity nLWhether n is metL≤25/(Yfin+ 1) condition:
If it is satisfied, then determining that center pixel i divides pixel for mistake, and mistake is divided to the label class label L of pixeliIt is modified in neighborhood
The most label class label of label class label quantity of the same race executes (8c);
If conditions are not met, then determining that center pixel i is correct distribution pixel, execute (8c);
(8c) judges label matrix ULIn pixel whether all traverse, if so, exporting final label matrix UF, it executes (8d),
Continue to traverse if it is not, returning to (8b);
(8d) is to final label matrix UFMiddle label class label is YWPixel assign different corresponding gray value xW, and will assign
The matrix of different gray values is exported as segmented image.
2. the method according to claim 1, wherein being used in (3) based on switch mean value constructing tactics to noise
The intuitionistic Fuzzy Sets of robustIt is accomplished by
(3a) first determines whether the grey scale pixel value multiplied in 3 neighborhoods the 3 of ith pixel is not all 0 or 255:
If set up, the pixel that gray value is 0 and 255 is taken out, then take the mean value of rest of pixels gray value, and be set as center
The amendment numerical value of pixel executes (3b);
If invalid, center pixel gray value is not modified, is executed (3b);
(3b) judges whether the pixel in input picture all traverses: if so, output amendment image, executes (3c), if not
It is to return to (3a) to continue to traverse;
The grey scale pixel value for correcting image is calculated as intuitionistic Fuzzy Sets by (3c)
(3c1) sets μ (xi) indicate grey scale pixel value xiDegree of membership, v (xi) and π (xi) it is x respectivelyiIt is constructed using Yager operator
Non-affiliated degree and hesitation degree, calculation formula difference it is as follows:
π(xi)=1- μ (xi)-v(xi),
Wherein, 0≤μ (xi)≤1,0≤v (xi)≤1,0≤μ (xi)+v(xi)≤1, α=0.8 indicates Yager operator coefficient;
The result construction intuitionistic Fuzzy Sets that (3c2) is obtained according to (3c1)
3. the method according to claim 1, wherein the grey value profile of input picture is directed in (4), using most
Small square law carries out curve fitting, and corresponding gray level is selected to be accomplished by as the cluster centre value of initialization
Gray level image is set as 256 gray levels by (4a), and grey level range is [0,255], and the gray scale for drawing gray level image is straight
Fang Tu;
(4b) counts the grey value profile frequency f (l) of the gray level l on grey level histogram, and does seven rank least square method curves
Fitting, wherein 0≤l≤255;
(4c) seeks the first derivative f'(l of f (l)), compare f'(l) with f (l) before and after gray level first derivative f'(l-1) and f'
(l+1), as l=0 l=255, only judge f'(l) whether less or greater than 0:
As l=0, f'(l is judged) whether < 0 is true, if so, being exported gray level l=0 as peak point;
As 0 < l < 255, f'(l is judged)=0 and f'(l-1) > 0 and f'(l+1) whether < 0 true, if so, by gray scale
Grade l is exported as peak point;
As l=255, f'(l is judged) whether > 0 is true, if so, being exported gray level l=255 as peak point;
(4d) judges whether 256 gray levels all traverse:
If so, output peak point, executes (4e), continue to traverse if it is not, returning to (4c);
(4e) sorts to the peak point that (4d) is exported according to the sequence of intensity profile frequency values from big to small, counts the number of peak point
Mesh is Ymax, and the corresponding gray level l of peak point sorted from large to small is counted, obtaining gray scale value range isIt is cluster centre initial value that gray level, which is corresponded to numerical statistic, such asIf cluster centre initial value range set C isWherein, lYIt indicates in peak point intensity profile frequency values according in descending order, the
Gray level corresponding to Y peak point intensity profile frequency values.
4. the method according to claim 1, wherein calculating grey scale pixel value x under intuitionistic Fuzzy Sets in (5b)iIt arrives
Cluster centre ckIntuitionistic Fuzzy DistancesIt is accomplished by
(5b1) sets μ (ck) indicate cluster centre ckDegree of membership, v (ck) and π (ck) respectively indicate cluster centre ckUse Yager
The non-affiliated degree and hesitation degree of operator construction, calculation formula difference are as follows:
π(ck)=1- μ (ck)-v(ck),
Wherein, 0≤μ (ck)≤1,0≤v (ck)≤1,0≤μ (ck)+v(ck)≤1;
(5b2) calculates Intuitionistic Fuzzy Distances according to the result that (5b1) is obtained
Wherein, μ (xi)、v(xi) and π (xi) respectively indicate grey scale pixel value xiDegree of membership, non-affiliated degree and hesitation degree.
5. the method according to claim 1, wherein utilizing the local spatial information of pixel and combination in (5c)
The linear weighting function coefficient H that (5a) is obtainedirThe Intuitionistic Fuzzy Distances obtained with (5b)Calculate intuitionistic fuzzy mesh
Scalar functions J, formula are as follows:
Wherein, ukiIt is pixel xiTo cluster centre ckSubordinating degree function, m indicate FUZZY WEIGHTED index.
6. according to the method described in claim 5, it is characterized in that, using lagrange's method of multipliers from intuitionistic fuzzy objective function
Subordinating degree function u is obtained in JkiWith cluster centre ckIntuitionistic Fuzzy SetsIt is accomplished by
(5c1) introduces Lagrange multiplier λ in intuitionistic fuzzy objective function Ji, construct the target letter with Lagrange multiplier
Number L:
(5c2) calculates separately L for subordinating degree function uki, cluster centre ckIntuitionistic Fuzzy SetsWith Lagrange multiplier λi's
Partial derivativeWithAnd in partial derivativeUnder conditions of, by acquiring u as followskiWith
Wherein, k=1,2 ..., Y, i=1,2 ..., n, r=1,2 ..., ω2。
7. the method according to claim 1, wherein being classified the formula table of range index evaluation index GD in (7)
Up to as follows:
Wherein, ui1It indicates that ith pixel is maximum and is subordinate to angle value, ui2Indicate ith pixel it is second largest be subordinate to angle value.
8. the method according to claim 1, wherein the maximum membership grade principle in (8a), is according to Vk=argk
{max(uki) pixel i is divided into the maximum classification V of degree of membershipk。
9. the method according to claim 1, wherein to final label matrix U in (8d)FMiddle label class label is
YWPixel assign different corresponding gray value xW, formula is as follows:
xW=[255/YW]
Wherein, [] indicates to take internal integer, final label matrix UFInside contain only categorization values 1,2 ..., YW,…,Yfin}。
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910507275.7A CN110211126B (en) | 2019-06-12 | 2019-06-12 | Image segmentation method based on intuitive fuzzy C-means clustering |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910507275.7A CN110211126B (en) | 2019-06-12 | 2019-06-12 | Image segmentation method based on intuitive fuzzy C-means clustering |
Publications (2)
Publication Number | Publication Date |
---|---|
CN110211126A true CN110211126A (en) | 2019-09-06 |
CN110211126B CN110211126B (en) | 2022-06-03 |
Family
ID=67792371
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910507275.7A Active CN110211126B (en) | 2019-06-12 | 2019-06-12 | Image segmentation method based on intuitive fuzzy C-means clustering |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110211126B (en) |
Cited By (12)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110634141A (en) * | 2019-09-19 | 2019-12-31 | 南京邮电大学 | Image segmentation method based on improved intuitionistic fuzzy c-means clustering and storage medium |
CN110956637A (en) * | 2019-12-02 | 2020-04-03 | 兰州交通大学 | Adaptive fuzzy C-means clustering noise image segmentation method and system |
CN111402272A (en) * | 2020-03-19 | 2020-07-10 | 西安石油大学 | Neighborhood weighted fuzzy clustering image segmentation method |
CN111476303A (en) * | 2020-04-09 | 2020-07-31 | 国网河北省电力有限公司电力科学研究院 | Line loss analysis method of fuzzy C-means clustering based on quantum optimization particle swarm |
CN111539966A (en) * | 2020-04-16 | 2020-08-14 | 华北电力大学 | Colorimetric sensor array image segmentation method based on fuzzy c-means clustering |
CN111932578A (en) * | 2020-08-12 | 2020-11-13 | 河北地质大学 | Pixel block clustering nondestructive testing image segmentation method integrating class index suppression factors |
CN112541528A (en) * | 2020-12-02 | 2021-03-23 | 国家电网有限公司 | Power transmission and transformation project cost prediction index optimization method based on fuzzy clustering |
CN112733825A (en) * | 2020-12-18 | 2021-04-30 | 北京城市排水集团有限责任公司 | Method for extracting image defect space of swing arm radar pipeline |
CN113269796A (en) * | 2021-06-10 | 2021-08-17 | 河北地质大学 | Image segmentation method and device and terminal equipment |
CN113409335A (en) * | 2021-06-22 | 2021-09-17 | 西安邮电大学 | Image segmentation method based on strong and weak joint semi-supervised intuitive fuzzy clustering |
CN114119503A (en) * | 2021-11-05 | 2022-03-01 | 华南师范大学 | Retina blood vessel segmentation method and device based on IFCM clustering |
CN115131588A (en) * | 2022-06-13 | 2022-09-30 | 西北工业大学 | Image robust clustering method based on fuzzy clustering |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108062757A (en) * | 2018-01-05 | 2018-05-22 | 北京航空航天大学 | It is a kind of to utilize the method for improving Intuitionistic Fuzzy Clustering algorithm extraction infrared target |
WO2018119808A1 (en) * | 2016-12-29 | 2018-07-05 | 浙江工商大学 | Stereo video generation method based on 3d convolutional neural network |
WO2019000455A1 (en) * | 2017-06-30 | 2019-01-03 | 上海联影医疗科技有限公司 | Method and system for segmenting image |
CN109145921A (en) * | 2018-08-29 | 2019-01-04 | 江南大学 | A kind of image partition method based on improved intuitionistic fuzzy C mean cluster |
CN109741330A (en) * | 2018-12-21 | 2019-05-10 | 东华大学 | A kind of medical image cutting method of mixed filtering strategy and fuzzy C-mean algorithm |
-
2019
- 2019-06-12 CN CN201910507275.7A patent/CN110211126B/en active Active
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2018119808A1 (en) * | 2016-12-29 | 2018-07-05 | 浙江工商大学 | Stereo video generation method based on 3d convolutional neural network |
WO2019000455A1 (en) * | 2017-06-30 | 2019-01-03 | 上海联影医疗科技有限公司 | Method and system for segmenting image |
CN108062757A (en) * | 2018-01-05 | 2018-05-22 | 北京航空航天大学 | It is a kind of to utilize the method for improving Intuitionistic Fuzzy Clustering algorithm extraction infrared target |
CN109145921A (en) * | 2018-08-29 | 2019-01-04 | 江南大学 | A kind of image partition method based on improved intuitionistic fuzzy C mean cluster |
CN109741330A (en) * | 2018-12-21 | 2019-05-10 | 东华大学 | A kind of medical image cutting method of mixed filtering strategy and fuzzy C-mean algorithm |
Non-Patent Citations (2)
Title |
---|
S.V. ARUNA KUMAR: ""A Modified Intuitionistic Fuzzy Clustering Algorithm for Medical Image Segmentation"", 《DE GRUYTER》 * |
赵凤: ""显著信息引导的直觉空间模糊聚类图像分割"", 《西安邮电大学学报》 * |
Cited By (17)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110634141B (en) * | 2019-09-19 | 2022-02-11 | 南京邮电大学 | Image segmentation method based on improved intuitionistic fuzzy c-means clustering and storage medium |
CN110634141A (en) * | 2019-09-19 | 2019-12-31 | 南京邮电大学 | Image segmentation method based on improved intuitionistic fuzzy c-means clustering and storage medium |
CN110956637A (en) * | 2019-12-02 | 2020-04-03 | 兰州交通大学 | Adaptive fuzzy C-means clustering noise image segmentation method and system |
CN111402272A (en) * | 2020-03-19 | 2020-07-10 | 西安石油大学 | Neighborhood weighted fuzzy clustering image segmentation method |
CN111402272B (en) * | 2020-03-19 | 2023-09-29 | 西安石油大学 | Neighborhood weighted fuzzy clustering image segmentation method |
CN111476303A (en) * | 2020-04-09 | 2020-07-31 | 国网河北省电力有限公司电力科学研究院 | Line loss analysis method of fuzzy C-means clustering based on quantum optimization particle swarm |
CN111539966A (en) * | 2020-04-16 | 2020-08-14 | 华北电力大学 | Colorimetric sensor array image segmentation method based on fuzzy c-means clustering |
CN111932578B (en) * | 2020-08-12 | 2022-06-07 | 河北地质大学 | Pixel block clustering nondestructive testing image segmentation method integrating class index suppression factors |
CN111932578A (en) * | 2020-08-12 | 2020-11-13 | 河北地质大学 | Pixel block clustering nondestructive testing image segmentation method integrating class index suppression factors |
CN112541528A (en) * | 2020-12-02 | 2021-03-23 | 国家电网有限公司 | Power transmission and transformation project cost prediction index optimization method based on fuzzy clustering |
CN112733825A (en) * | 2020-12-18 | 2021-04-30 | 北京城市排水集团有限责任公司 | Method for extracting image defect space of swing arm radar pipeline |
CN113269796A (en) * | 2021-06-10 | 2021-08-17 | 河北地质大学 | Image segmentation method and device and terminal equipment |
CN113269796B (en) * | 2021-06-10 | 2023-08-25 | 河北地质大学 | Image segmentation method and device and terminal equipment |
CN113409335A (en) * | 2021-06-22 | 2021-09-17 | 西安邮电大学 | Image segmentation method based on strong and weak joint semi-supervised intuitive fuzzy clustering |
CN114119503A (en) * | 2021-11-05 | 2022-03-01 | 华南师范大学 | Retina blood vessel segmentation method and device based on IFCM clustering |
CN115131588A (en) * | 2022-06-13 | 2022-09-30 | 西北工业大学 | Image robust clustering method based on fuzzy clustering |
CN115131588B (en) * | 2022-06-13 | 2024-02-23 | 西北工业大学 | Image robust clustering method based on fuzzy clustering |
Also Published As
Publication number | Publication date |
---|---|
CN110211126B (en) | 2022-06-03 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110211126A (en) | Image partition method based on intuitionistic fuzzy C mean cluster | |
CN108665120B (en) | Method and device for establishing scoring model and evaluating user credit | |
CN105741279B (en) | The quick image partition method for suppressing fuzzy clustering based on rough set | |
CN108288033B (en) | A kind of safety cap detection method based on random fern fusion multiple features | |
CN101556650B (en) | Distributed self-adapting pulmonary nodule computer detection method and system thereof | |
CN103761726B (en) | Block adaptive image partition method based on FCM | |
Zhou et al. | Salient object detection via fuzzy theory and object-level enhancement | |
CN109145921A (en) | A kind of image partition method based on improved intuitionistic fuzzy C mean cluster | |
WO2020150896A1 (en) | Data association method and apparatus for video target, and storage medium | |
CN103914705B (en) | Hyperspectral image classification and wave band selection method based on multi-target immune cloning | |
CN110569782A (en) | Target detection method based on deep learning | |
CN106203521B (en) | The SAR image change detection learnt based on disparity map from step | |
CN106991686B (en) | A kind of level set contour tracing method based on super-pixel optical flow field | |
CN105740915B (en) | A kind of collaboration dividing method merging perception information | |
CN109684922A (en) | A kind of recognition methods based on the multi-model of convolutional neural networks to finished product dish | |
CN110188763B (en) | Image significance detection method based on improved graph model | |
CN109801305B (en) | SAR image change detection method based on deep capsule network | |
Sun et al. | Nearest neighbors-based adaptive density peaks clustering with optimized allocation strategy | |
CN117197140B (en) | Irregular metal buckle forming detection method based on machine vision | |
Wang et al. | A spectral clustering method with semantic interpretation based on axiomatic fuzzy set theory | |
CN106600610B (en) | FCM image segmentation method and device | |
CN108846845B (en) | SAR image segmentation method based on thumbnail and hierarchical fuzzy clustering | |
CN106056165A (en) | Saliency detection method based on super-pixel relevance enhancing Adaboost classification learning | |
Wu et al. | Cloud model-based method for range-constrained thresholding | |
CN107564008B (en) | Rapid SAR image segmentation method based on key pixel fuzzy clustering |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |