how to implement code of cluster based segmentation on image?

Question

Prerna Surbhi 2016-4-20

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/279995-how-to-implement-code-of-cluster-based-segmentation-on-image

编辑： Prerna Surbhi 2016-4-20

I have a kmean clutering for segmenting any gray or color image,i want to know what changes i have to do to implement it in any image .please help me to sort out?

function [lb,center] = adaptcluster_kmeans(im)

% This code is written to implement kmeans clustering for segmenting any % Gray or Color image. There is no requirement to mention the number of cluster for % clustering. % IM - is input image to be clustered. % LB - is labeled image (Clustered Image). % CENTER - is array of cluster centers. % Execution of this code is very fast. % It generates consistent output for same image.

% Written by Ankit Dixit. % January-2014.

if size(im,3)>1 [lb,center] = ColorClustering(im); % Check Image is Gray or not. else [lb,center] = GrayClustering(im); end

function [lb,center] = GrayClustering(gray) gray = double(gray); array = gray(:); % Copy value into an array. % distth = 25; i = 0;j=0; % Intialize iteration Counters. tic while(true) seed = mean(array); % Initialize seed Point. i = i+1; %Increment Counter for each iteration. while(true) j = j+1; % Initialize Counter for each iteration. dist = (sqrt((array-seed).^2)); % Find distance between Seed and Gray Value. distth = (sqrt(sum((array-seed).^2)/numel(array)));% Find bandwidth for Cluster Center. % distth = max(dist(:))/5; qualified = dist<distth;% Check values are in selected Bandwidth or not. newseed = mean(array(qualified));% Update mean.

        if isnan(newseed) % Check mean is not a NaN value.
            break;
        end
        if seed == newseed || j>10 % Condition for convergence and maximum iteration.
            j=0;
            array(qualified) = [];% Remove values which have assigned to a cluster.
            center(i) = newseed; % Store center of cluster.
            break;
        end
        seed = newseed;% Update seed.
    end
    if isempty(array) || i>10 % Check maximum number of clusters.
        i = 0; % Reset Counter.
        break;
    end

end toc

center = sort(center); % Sort Centers. newcenter = diff(center);% Find out Difference between two consecutive Centers. intercluster = (max(gray(:)/10));% Findout Minimum distance between two cluster Centers. center(newcenter<=intercluster)=[];% Discard Cluster centers less than distance.

% Make a clustered image using these centers.

vector = repmat(gray(:),[1,numel(center)]); % Replicate vector for parallel operation. centers = repmat(center,[numel(gray),1]);

distance = ((vector-centers).^2);% Find distance between center and pixel value. [~,lb] = min(distance,[],2);% Choose cluster index of minimum distance. lb = reshape(lb,size(gray));% Reshape the labelled index vector.

function [lb,center] = ColorClustering(im)

im = double(im); red = im(:,:,1); green = im(:,:,2); blue = im(:,:,3);

array = [red(:),green(:),blue(:)]; % distth = 25; i = 0;j=0; tic while(true)

    seed(1) = mean(array(:,1));
    seed(2) = mean(array(:,2));
    seed(3) = mean(array(:,3));
    i = i+1;
    while(true)
        j = j+1;
        seedvec = repmat(seed,[size(array,1),1]);
        dist = sum((sqrt((array-seedvec).^2)),2);
         distth = 0.25*max(dist);
        qualified = dist<distth;
        newred = array(:,1);
        newgreen = array(:,2);
        newblue = array(:,3);
        newseed(1) = mean(newred(qualified));
        newseed(2) = mean(newgreen(qualified));
        newseed(3) = mean(newblue(qualified));
        if isnan(newseed)
            break;
        end
        if (seed == newseed) | j>10
            j=0;
            array(qualified,:) = [];
            center(i,:) = newseed;
            %             center(2,i) = nnz(qualified);
            break;
        end
        seed = newseed;
    end
    if isempty(array) || i>10
        i = 0;
        break;
    end

end toc centers = sqrt(sum((center.^2),2)); [centers,idx]= sort(centers);

while(true) newcenter = diff(centers); intercluster =25; %(max(gray(:)/10)); a = (newcenter<=intercluster); % center(a,:)=[]; % centers = sqrt(sum((center.^2),2)); centers(a,:) = []; idx(a,:)=[]; % center(a,:)=0; if nnz(a)==0 break; end

end center1 = center; center =center1(idx,:); % [~,idxsort] = sort(centers) ; vecred = repmat(red(:),[1,size(center,1)]); vecgreen = repmat(green(:),[1,size(center,1)]); vecblue = repmat(blue(:),[1,size(center,1)]);

distred = (vecred - repmat(center(:,1)',[numel(red),1])).^2; distgreen = (vecgreen - repmat(center(:,2)',[numel(red),1])).^2; distblue = (vecblue - repmat(center(:,3)',[numel(red),1])).^2;

distance = sqrt(distred+distgreen+distblue); [~,label_vector] = min(distance,[],2); lb = reshape(label_vector,size(red)); %