Error:Empty cluster at iteration

Finally for the question in

http://www.mathworks.in/matlabcentral/answers/55028-finding-mean-for-values

i have found mean and for k means i get error

my code

By this way i have found mean

for i2=1:192

    kl(:,:,i2)=kkk(:,:,i2);

    kll=kkk(:,:,i2);

    k3(:,:,i2)=mean(kll,3);

end

next i tried to do k means

for i2=1:192

    k4(:,:,i2)=k3(:,:,i2);

    k5=k4(:);

    k3(:,:,i2)=kmeans(k5,3);
end

I get error

??? Error using ==> kmeans>batchUpdate at 435 Empty cluster created at iteration 1.

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

Error in ==> Samp2 at 52 k3(:,:,i2)=kmeans(kll,3);

please assist

[Jan]

After kll=kkk(:,:,i2), the variable k11 is a matrix. Then using mean(k11, 3) does not compute anything, but replies k11. Finally k3 should be the same as kkk and therefore I do not understand what you are doing.

Answer 1

[Walter Roberson]

"empty cluster" can occur by chance (kmeans does random initialization of positions), but more likely is the possibility that your points naturally form fewer than 3 clusters.

[FIR]

ok walter thanks another thing is that i need to find the nearby values with difference around 5 (the difference if calculated ny finding their mean)and find the mean valued of them

[Walter Roberson]

What does "nearby" mean in this context. What does "around 5" mean in this context? How does this relate to using kmeans() ?

[FIR]

this does not relates to kmeans...in that i have found mean,,for all

now i want to check which al values are nearby ,at a difference of 5 and for that we have to find mean value

suppose val(:,:,1) has mean value of 145.23 and val(:,:,56) has mean of 149.10and val(:,:,103) has mean of 143.05 for all these three we have to find the mean,and so on

[Walter Roberson]

This current Question is about kmeans. Please move the other topic to an appropriate question.

Answer 2

[ishaan]

I received the above error too.. I am applying kmeas clustering on a Canny's Edge detected image with k param as 2.

I3 = imclose(I2,strel('disk',20));
Imed = medfilt2(rgb2gray(I3));
Ifilt = edge(Imed,'canny');

%I3 = applycform(Ifilt,makecform('srgb2lab'));
%I = double(I3(:,:,2));
I = double(Ifilt(:,2));
szx2 = size(I,1);
szy2 = size(I,2);

I = reshape(I,szx2*szy2,1);
nColors = 2;
%%Image Manipulation Kmeans...............................................

[idx ct] = kmeans(I,nColors);

However i Received the empty cluster error. How to avoid the error. I need the edge detected image as well.