applying colour for clustered image

i have a code below of 2 images

for n=1:length(dirlist);
    img = imread([pathname, dirlist(n).name]);

I=img;
HSV = rgb2hsv(I);
%HSV=I;
H = HSV(:,:,1); H = H(:);
S = HSV(:,:,2); S = S(:);
V = HSV(:,:,3); V = V(:);
idx = kmeans([H S V],3);
figure,imshow(ind2rgb(reshape(idx, size(I,1), size(I, 2)), [0 0 1; 0 0.8 0;1 0 0]))
end

the image consist of water,and land

one image consist of 70% of water and other 40% of water is it possible to give same colour for water in two images

plz assist

Answer 1

[ChristianW]

You can define the water color. Than calculate the mean color for every cluster. Compare the mean cluster color with your defined water color to find your water cluster.

You could also use the mean cluster color directly as cluster color.

[kash]

Thanks Christia can you plz tell what is mean cluster color

[ChristianW]

Mean [Red Green Blue] values for all pixels in a cluster.

Or use rgb2ind with nColor = 1:

RGB = imread('ngc6543a.jpg');
[X,map] = rgb2ind(RGB, 1); % map is the mean color
imtool(permute(map,[3 1 2]))

[kash]

christain what must be done with this code,can you please tell in brief

My code for mean value for each cluster,plz tell what must be done next

HSV = rgb2hsv(I);
H = HSV(:,:,1); H = H(:);
S = HSV(:,:,2); S = S(:);
V = HSV(:,:,3); V = V(:);
idx = kmeans([H S V], 3);
% imshow(ind2rgb(reshape(idx, size(I,1), size(I, 2)), [0 0 1; 0 0.8 0]))
cluster_idx=idx;

pixel_labels = reshape(cluster_idx,size(I,1),size(I,2));
figure,imshow(pixel_labels,[]), title('image labeled by cluster index');
segmented_images = cell(1,3);
rgb_label = repmat(pixel_labels,[1 1 3]);

for k = 1:3
    color = I;
    color(rgb_label ~= k) = 0;
    segmented_images{k} = color;
end

figure,imshow(segmented_images{1}), title('objects in cluster 1');

figure,imshow(segmented_images{2}), title('objects in cluster 2');

figure,imshow(segmented_images{3}), title('objects in cluster 2');

Mean1=mean(mean(segmented_images{1}))

Mean2=mean(mean(segmented_images{2}))

Mean3=mean(mean(segmented_images{3}))

[Image Analyst]

Where can we find 'ngc6543a.jpg' so we can run this?

[kash]

My image consists of water and land

[ChristianW]

ngc6543a.jpg is a Matlab example image, but I guess IA is communicating the helpfulness of having access to your image.

You falsifies the mean values with a lot of black color. This is what I mean:

I = imread('ngc6543a.jpg');
HSV = I; %rgb2hsv(I);
H = HSV(:,:,1); H = H(:);
S = HSV(:,:,2); S = S(:);
V = HSV(:,:,3); V = V(:);
Y = double([H S V]);
nColors = 3;
idx = kmeans(Y,nColors,'Replicates',3);

% get the mean cluster colors
pixel_labels = reshape(idx,size(I,1),size(I,2)); % imshow(pixel_labels,[]);
[X,map] = rgb2ind(I,65536); % indexed original image
for i = 1:nColors
    c(i,:) = mean(map(X(idx==i)+1,:)); %#ok<SAGROW>
end
imshow(pixel_labels,c)

% compare with Water color
Nova = [1 0 0]; % Nova color or Water
[C,Icmin] = min(sum((c-ones(size(c,1),1)*Nova).^2,2)); % distance
c(Icmin,:) = Nova;
imshow(pixel_labels,c)

As a note, kmeans crashed sometimes here, but thats your part of the code or it might be the example image.

[Image Analyst]

'ngc6543a.jpg' is not one of the official demo images that come with the Image Processing Toolbox, and located in folder C:\Program Files\MATLAB\R2012b\toolbox\images\imdemos. I don't know how/where to obtain the image.

[ChristianW]

Its a Matlab image (not IPToolbox), at least in my version. Its still just an random image I took to run his code. But whatever, here you go: ngc6543a.jpg

[kash]

Christian thanks a lot for showing interest in my project ,I have two images consisting of water and land

http://imgur.com/1Z4PH2P,EjQhlVo#1

i tried your code,but the water portion d assigned with different colours

kindly assist

[ChristianW]

You are messing things up. I took your color/texture segmentation (implemented with kmeans), the segmentation isn't my part of the code. My part just takes a kmeans output and does a cluster affectation. It applies color for the clustered image, thats what you've asked for.

If your question is aimed at improving your segmentation, than Image Analyst is your best bet.

[kash]

my question also deals with this,having two images consisting of water,how can we give same colour for water for different images

Answer 2

[Image Analyst]

Looks like you can find the blue water fairly easily with color classification. The green water you'll have to find by adding in a texture filter like stdfilt() or entropyfilt() because the trees are very close to the same color and the only difference is the texture. The white stuff is either sand or surf on sand - either way white stuff is land. If you still need help, say so. I have color classification examples in my File Exchange http://www.mathworks.com/matlabcentral/fileexchange/?term=authorid%3A31862

[kash]

I am stuck here for bit long,can u help me in giving an example please ,as per you instruction i did as

    I=imread('water2.jpg');

    d=I(:,:,1);
    d1=I(:,:,2);
    d2=I(:,:,3);

    d=d(:);
    d1=d1(:);
    d2=d2(:);

    d=stdfilt(d);
    d1=stdfilt(d1);
    d2=stdfilt(d2);

     idx=kmeans([d d1 d2],3);
     imshow(ind2rgb(reshape(idx, size(I,1), size(I, 2)), [0 0 1; 0 0.8 0]))

now how we can appply same colour for different images consisting of water

kindly assist

[Walter Roberson]

Please stop indenting your text when you post. Only indent your code. I have fixed your posting yet again.

[Walter Roberson]

In the image with the trees, I cannot tell whether the portion to the upper right is water (making the portion with trees either a spit of land or a portion of a lagoon), or if that portion is instead grass or something similar.

[kash]

Actuall walter colouring must be mase with more or less same properties ,

is it possible

[kash]

from the above two images is it possible to say ,each image consists of how much percentahe of water for each image,(only water)

[Walter Roberson]

So in that image with the trees, is the upper right portion calm water or is it vegetation ?

[kash]

its water

[Image Analyst]

Exactly what does "Walter colouring must be mase" mean?

[kash]

I cannot understand what u said,what is walter colouring