See here, i think it can help you :

http://www.mathworks.com/matlabcentral/fileexchange/10161-mean-shift-clustering

See here, i think it can help you :

http://www.mathworks.com/matlabcentral/fileexchange/10161-mean-shift-clustering

Dear @Aymen it's for mean shift calculation, not for segmentation, so please give me idea about segmentation.

I'm sure there is a code on Matlab provided on net but I didn't find it on my search.

But this one seems similar, maybe a lead

http://stackoverflow.com/questions/4831813/image-segmentation-using-mean-shift-explained

function [origDataPoints,dataPoints,clusterCentroids,pointsToClusters] = meanshift(numDimensions,useKNNToGetH)

H = 1; %if useKNNToGetH is false, H will be used else knn will be used to determine H

threshold = 1e-3;

numPoints = 300;

k = 100; % when we use knn this is the k that is used.

numClasses = 0;

pointsToCluster2 = [];

centroids = [];

function[dataPoints] = getDataPoints(numPoints,numDimensions)

numClasses = randi([2,8],1); % generate a random number between 2 and 8

dataPoints = [];

curNumberOfPoints = 0;

randomMeans = [];

for i = 1:numClasses

randomMean = randi([0,100], 1) * rand(1);

randomStd = randi([1,2]) * rand(1);

curGaussianModelPoints = randomMean + randomStd .* randn(numPoints,numDimensions);

dataPoints = [dataPoints;curGaussianModelPoints];

randomMeans = [randomMeans,randomMean];

pointsToCluster2(curNumberOfPoints +1 : (curNumberOfPoints+numPoints)) = i;

curNumberOfPoints = curNumberOfPoints + numPoints;

centroids = randomMeans;

end

end

function d = sqdist(a,b)

%taken from demo code of class

aa = sum(a.*a,1); bb = sum(b.*b,1); ab = a'*b;

d = abs(repmat(aa',[1 size(bb,2)]) + repmat(bb,[size(aa,2) 1]) - 2*ab);

end

function bandwidth = getBandWith(curDataPoint,origDataPoints,euclideanDist,useKNNToGetH)

if (useKNNToGetH == false)

bandwidth = H;

else

[sortedEuclideanDist,indexInOrig] = sort(euclideanDist);

bandwidth = sortedEuclideanDist(k);

end

end

function [clusterCentroids,pointsToCluster] = getClusters(origDataPoints,finalDataPoints)

[numSamples,numFeatures] = size(finalDataPoints);

clusterCentroids = [];

pointsToCluster = [];

numClusters = 0;

for i = 1:numSamples

closestCentroid = 0 ;

curDataPoint = finalDataPoints(i,:);

for j = 1:numClusters

distToCentroid = sqrt(sqdist(curDataPoint',clusterCentroids(j,:)'));

%distToCentroid = sqdist(curDataPoint',clusterCentroids');

%if (distToCentroid < 8 * H)

if (distToCentroid < 4 * H)

closestCentroid = j;

break;

end

end

if (closestCentroid > 0)

pointsToCluster(i,:) = closestCentroid;

clusterCentroids(closestCentroid,:) = 0.5 * (curDataPoint + clusterCentroids(closestCentroid,:));

else

numClusters = numClusters + 1 ;

clusterCentroids(numClusters,:) = finalDataPoints(i,:);

pointsToCluster(i,:) = numClusters;

end

end

end

function plotPoints(clusterCentroids,pointsToCluster,origDataPoints)

[numSamples,numFeatures] = size(origDataPoints);

allColorsInMatlab = 'bgrcmyk'; %ignoring white as it is the default bg for the plot

sizeOfColors = size(allColorsInMatlab,2);

if (numFeatures == 2)

% plot the original Data Points

h = figure(1);

hold on;

for i=1:numClasses

colourIndex = mod(i,sizeOfColors) + 1;

allElemsInThisCluster = find(pointsToCluster2 == i);

allOrigPointsInThisCluster = origDataPoints(allElemsInThisCluster,:);

plot(allOrigPointsInThisCluster(:,1), allOrigPointsInThisCluster(:,2), [allColorsInMatlab(colourIndex) '.']);

end

plot(centroids,centroids,'s');

hold off;

h = figure(2);

hold on;

% plot the original Data Points

for i = 1:size(clusterCentroids,1)

colourIndex = mod(i,sizeOfColors) + 1 ;

allElemsInThisCluster = find(pointsToCluster == i);

allOrigPointsInThisCluster = origDataPoints(allElemsInThisCluster,:);

plot(allOrigPointsInThisCluster(:,1), allOrigPointsInThisCluster(:,2), [allColorsInMatlab(colourIndex) '.']);

end

hold off;

end

end

function [origDataPoints,dataPoints] = doMeanShift(dataPoints,useKNNToGetH)

[numSamples,numFeatures] = size(dataPoints);

origDataPoints = dataPoints;

for i = 1:numSamples

diffBetweenIterations = 10;

while (diffBetweenIterations > threshold)

curDataPoint = dataPoints(i,:);

euclideanDist = sqdist(curDataPoint',origDataPoints');

bandwidth = getBandWith(origDataPoints(i,:),origDataPoints,euclideanDist,useKNNToGetH);

kernelDist = exp(-euclideanDist ./ (bandwidth^2));

numerator = kernelDist * origDataPoints;

denominator = sum(kernelDist);

newDataPoint = numerator/denominator;

dataPoints(i,:) = newDataPoint;

diffBetweenIterations = abs(curDataPoint - newDataPoint);

end

end

[clusterCentroids,pointsToClusters] = getClusters(origDataPoints,dataPoints);

plotPoints(clusterCentroids,pointsToClusters,origDataPoints);

end

dataPoints = getDataPoints(numPoints,numDimensions);

[origDataPoints,dataPoints] = doMeanShift(dataPoints,useKNNToGetH);

end

First thanks all for help

@Sander, This code is for Meanshift only and I already have this code, I require a code for Mean shift segmentation.

Still I don't have my correct solution

Once you have detected the centroid through the mean shift you can easily segment it through regionprop command of matlab

@Sander, can you please more elaborate it.

in the above code you are detecting centroids and clustering it...mean shift algorithm is implemented via the above code now segmentation part can be done by regionprops where you can give the BW image and the centroid which you have detected with the help of mean shift algorithm, I suggest you should see regionprops and watershed from matlab help you will understand my suggestion, atleast you can give it a try though

.

the link below might be of interest to you :

Sylvain Arlot, Alain Celisse. Segmentation of the mean of heteroscedastic data via cross-validation. Statistics and Computing (2010)

(with matlab code ; note the _heteroscedastic_ setting !)

http://math.univ-lille1.fr/~celisse/softwares.html

in the same heteroscedastic spirit :

Alexis Bondu, Marc Boullé: A supervised approach for change detection in data streams. IJCNN 2011: 519-526

(no code provided but relies on a tool available as shareware)

.

Dear Sandeep Singh Sengar, Below is the link. Check may be of your use.

http://www.mathworks.com/matlabcentral/fileexchange/10161-mean-shift-clustering

@Afaq above link program is for same purpose to find clustering and meanshift as given by Sander Ali

To understand mean shift segmentation, check following publication

Comaniciu, Dorin, and Peter Meer. "Mean shift: A robust approach toward feature space analysis." Pattern Analysis and Machine Intelligence, IEEE Transactions on 24, no. 5 (2002): 603-619.

&

Publicly available code is attached as zip file

Mean Shift Segmentation relevant links:

http://www.shawnlankton.com/2007/11/mean-shift-segmentation-in-matlab/

http://www.wisdom.weizmann.ac.il/~bagon/matlab.html

http://coewww.rutgers.edu/riul/research/code/EDISON/index.html

http://www.wisdom.weizmann.ac.il/~bagon/matlab_code/edison_matlab_interface.tar.gz