clustering based on area

i already measured the area on my sputum cell. now i want to cluster it based on area. area >2000 in figure 1 and area 2000< is in other figure 2. can i know the simple tutorial.? i dont want to use k mean clustering because it very complicated. i only want to cluster them as simple as i can

[martin SALGADO]

Might uploading your code. I'm doing my memory on this topic and I would be very useful your code. Thank you

[Image Analyst]

You can look at my Image Segmentation Tutorial here: http://www.mathworks.com/matlabcentral/fileexchange/?term=authorid%3A31862

No need to do it inefficiently like wil did below - you can simply use ismember() like I did.

[martin SALGADO]

Your code Image Segmentation Tutorial is for segment images, but I want only one way clustering a set of points, and the area that contains the items is less than some value.

[izyan hanum]

you can email me [e-mail]

Answer 1

[wil]

Hi, I've looked at the code and I've worked out what you need. Before your final loop "% Loop over all blobs printing their measurements to the command window", add the line

vols = zeros(1,numberOfBlobs); % create list to contain cell areas

and then at the end of the loop (before end), simply add

vols(k) = blobArea; % add area to vols

Then, the code you require to plot two images (one for the smaller, one for the bigger) as binary images. The lists idxs_1 and idx_2 contain the indexes for the cells belonging to each group. You can use this to add more qualifiers, and create different groups if you need to.

idxs_1 = find(vols < 2000);  % indexes of vols that are lower than 2000 
idxs_2 = find(vols >= 2000);

cells_1 = zeros(size(hImage)); % create blank binary images
cells_2 = zeros(size(hImage));

for i = 1:numel(idxs_1) % add small blobs to image 1
    cells_1(blobMeasurements(idxs_1(i)).PixelIdxList) = 1;
end
for i = 1:numel(idxs_2) % add larger blobs to image 2
    cells_2(blobMeasurements(idxs_2(i)).PixelIdxList) = 1;
end
% display
figure(4), subplot (2,2,1), imshow(cells_1,[]);
figure(4), subplot (2,2,2), imshow(cells_2,[]);

I hope this solves your problem.

Cheers, Wil

[izyan hanum]

yes sir.............. that i want sir..... im very happy now... thank you may god bless you..

[Image Analyst]

It's simpler to just use bwareafilt().

Answer 2

[wil]

Hi,

k-means is probably the simplest method of clustering, but if you simply want to sort the regions based on their area, you can use logical indexing or the find() method.

It depends on they type of data your groups are, but if they are simply binary images you can use (assuming the cells are contained in a cell, change as appropriate)

   vol = cells{i};
   area(i) = numel(vol(vol ~= 0));

to get the area for each cell and sort them using

   idxs_1 = find(area < 2000);
   idxs_2 = find(area >= 2000);
   cells_1 = cells(idxs_1);
   cells_2 = cells(idxs_2);

The groups cells_1 and cells_2 now contain your grouped cells.

Hope this helps, let me know if anything needs clarifying.

Wil

[izyan hanum]

i run but still have error :( this is my picture

Answer 3

[izyan hanum]

i dont know how to send the code here using microsoft

[izyan hanum]

https://word.office.live.com/wv/WordView.aspx?FBsrc=https%3A%2F%2Fwww.facebook.com%2Fattachments%2Ffile_preview.php%3Fid%3D734274853338438%26time%3D1427127122%26metadata&access_token=100003736390577%3AAVKwfhY8Bg_OXG_qYVHRh4hByGWdBCkH8QqLBSW3UOoxzQ&title=Try+thresholding+in+HSV+color+space+to+get+purple+blobs+but+not+black+blobs+or+yellow+background.docx