K-means for a grayscale image

I've tried the below code to cluster the grayscale image,

I = imread('sym_059.tif');
I = im2double(I);
c = kmeans(I, 3);
p = reshape(c, size(I));

executing this code, I am getting error as follows "Error using reshape-To RESHAPE the number of elements must not change." How can I debug this.? Help appreciated.

[gurleen ghuman]

i m getting this error . plz help... Error in ==> kmeans at 336 converged = batchUpdate();

Error in ==> Untitled5 at 3 c = kmeans(I(:), 3);

[Image Analyst]

Did you try this:

clc;    % Clear the command window.
clear all;
close all;
workspace;  % Make sure the workspace panel is showing.
format long g;
format compact;
fontSize = 24;

grayImage = double(imread('cameraman.tif'));
subplot(1, 2, 1);
imshow(grayImage, []);
title('Original Image', 'FontSize', fontSize);

numberOfClasses = 3;
indexes = kmeans(grayImage(:), numberOfClasses);
classImage = reshape(indexes, size(grayImage));

h2 = subplot(1, 2, 2);
imshow(classImage, []);
title('Classified Image', 'FontSize', fontSize);
colormap(h2, lines(numberOfClasses));
colorbar;

% Set up figure properties:
% Enlarge figure to full screen.
set(gcf, 'Units', 'Normalized', 'OuterPosition', [0 0 1 1]);
% Get rid of tool bar and pulldown menus that are along top of figure.
set(gcf, 'Toolbar', 'none', 'Menu', 'none');
% Give a name to the title bar.
set(gcf, 'Name', 'Demo by ImageAnalyst', 'NumberTitle', 'Off')

[mumu aktar]

I have tried this code but why it works for only 3 classes not more than this. when I take noOfclasses=5 the following error is shown: ??? Error using ==> kmeans>batchUpdate at 435 Empty cluster created at iteration 1.

Error in ==> kmeans at 336 converged = batchUpdate();

Error in ==> Untitled5 at 15 indexes = kmeans(grayImage(:), numberOfClasses);

[hmayda mounira]

i m getting this erreur

Error using reshape To RESHAPE the number of elements must not change. Error in expemple (line 22) classImage = reshape(indexes, size(grayImage));

please can I help you

[Walter Roberson]

hmayda mounira:

Did you possibly try to run this with an image that is not already grayscale? The demo image that Image Analyst used, cameraman.tif, is grayscale.

If you read in a JPEG image then even if it looks gray it almost certainly is RGB, so you might need to use rgb2gray() on your image.

[Iremsu Savas]

Hello, I tried the code you used and the final result of the image is just white all the time. Is there something that I skip to do? Can you please help me?

[Walter Roberson]

imshow(uint8(p))

[Image Analyst]

Try my code (hidden in the comments above), NOT the code that anusha says there is a problem with.

Answer 1

[Walter Roberson]

kmeans returns a vector of cluster indices, one index per row of input. You are trying to reshape that as if it had as many entries as the number of pixels in your image.

Possibly you want to try

c = kmeans(I(:), 3);

[anusha reddy]

It worked. Thank You !

[misbah khan]

hi i want to ask how to show resulted image after applying kmean to see output

[Walter Roberson]

Ic = reshape(c, size(I));
imshow(Ic,[])

[misbah khan]

I = imread('sym_059.tif');

I = im2double(I);

c = kmeans(I(:), 3);

Ic = reshape(c, size(I));

imshow(Ic,[])

it gives error in this code what was the problem here?

[misbah khan]

it gives error in k means line

[Image Analyst]

You probably don't have a license for the Stats toolbox. Do you see it listed then you do this:

>> ver

Have you tried my code:

clc;    % Clear the command window.
clear all;
close all;
workspace;  % Make sure the workspace panel is showing.
format long g;
format compact;
fontSize = 24;

grayImage = double(imread('cameraman.tif'));
subplot(1, 2, 1);
imshow(grayImage, []);
title('Original Image', 'FontSize', fontSize);

numberOfClasses = 3;
indexes = kmeans(grayImage(:), numberOfClasses);
classImage = reshape(indexes, size(grayImage));

h2 = subplot(1, 2, 2);
imshow(classImage, []);
title('Classified Image', 'FontSize', fontSize);
colormap(h2, lines(numberOfClasses));
colorbar;

% Set up figure properties:
% Enlarge figure to full screen.
set(gcf, 'Units', 'Normalized', 'OuterPosition', [0 0 1 1]);
% Get rid of tool bar and pulldown menus that are along top of figure.
set(gcf, 'Toolbar', 'none', 'Menu', 'none');
% Give a name to the title bar.
set(gcf, 'Name', 'Demo by ImageAnalyst', 'NumberTitle', 'Off')
fprintf('Done running %s.m.\n', mfilename);

[misbah khan]

hi iwant same simple code just want to ask you why it shows error?

[misbah khan]

hi apply your code still error below

indexes = kmeans(grayImage(:), numberOfClasses);

[misbah khan]

Error in newk (line 16)

[misbah khan]

i check >>ver

it returns 1

[Image Analyst]

Line 16 is

classImage = reshape(indexes, size(grayImage));

I don't even have a newk variable in any of my code.

Just to double check, I copied my demo code above and rand it and it ran flawlessly with no errors. So you must have changed something. You'd have to include your code, and the actual error message itself - you just gave the line that threw the unknown error. So I suspect you either did not assign numberOfClasses or assigned an invalid value to it. And same for the grayImage variable.

[misbah khan]

newk is my file name

[misbah khan]

check

[misbah khan]

have you check my error file?

[Walter Roberson]

You need to rename or delete /MATLAB Drive/kmeans.m as it is interfering with calling the kmeans function

[misbah khan]

Hi. Thank you, it's resolved.

One more question : how do I find means for array in a loop?

My code is here:

array[]
loop
arr=(array,m)
loop end
m = mean(arr)

It stores the last iteration mean value in m. How do I add all values when the loop iterates?

[Walter Roberson]

That is not MATLAB code.

Also, you are creating arr from array and m but if you wanted to accumulate results you would have to create output from arr and m . For example,

arr = [];
for K = 1 : 10
    m = randi(10);
    arr = [arr, m];
end

[misbah khan]

and how to find mean for these array values ?syntax?

[Walter Roberson]

mean(arr)

[Salma Hassan]

What about several images? How can I cluster them into k clusters?

[Walter Roberson]

Provided the images are the same number of pixels, and are all RGB or are all grayscale, then construct an array in which each row is reshape() of an image into a single row, and the rows correspond to different images. Then kmeans() .

This would attempt to cluster the images as a whole into clusters.

Answer 2

[Image Analyst]

Demos for kmeans for images attached.