bwImg = imread('Cattle.jpg');

[coords, lblImg] = bwboundaries(bwImg, 4, 'noholes');

x = coords{1}(1:end-1,1);

y = coords{1}(1:end-1,2);

xc = sum(x)/length(x);

yc = sum(y)/length(y);

rp = sqrt((x - xc).^2 + (y - yc).^2);

%This code assumes that you have only one object. Otherwise, you will get multiple list

% in 'coords' which have to be deal individually in the same manner 

I think the previous post calculates the average X and Y of contour-pixels, which is not equal to the centroid.

Use the matlab function regionprops(); much easier.

Thanks Lambert. You can get the centroid in the same way using segmented image instead of edge image. I am modifying the code:

bwImg = imread('Cattle.jpg');

linIndex = find(bwImg);

[xCoords, yCoords] = ind2sub(size(bwImg),linIndex);

xc = sum(xCoords)/length(xCoords);

yc = sum(yCoords)/length(yCoords);

[coords, lblImg] = bwboundaries(bwImg, 4, 'noholes');

x = coords{1}(1:end-1,1);

y = coords{1}(1:end-1,2);

rp = sqrt((x - xc).^2 + (y - yc).^2);

You can also use regionprops as below:

bwImg = imread('Cattle.jpg');

lbl = bwconncomp(bwImg);

stat = regionprops(lbl,'Centroid');

xc = stat(1).Centroid(2);

yc = stat(1).Centroid(1);

[coords, lblImg] = bwboundaries(bwImg, 4, 'noholes');

x = coords{1}(1:end-1,1);

y = coords{1}(1:end-1,2);

rp = sqrt((x - xc).^2 + (y - yc).^2);

Hi M. The average of the X and Y coordinates of all cow pixels equals the centroid. The average of the X and Y coordinates of only the cow edge pixels does not equal the centroid. Right?

Hi Lambert. You are right - centroid is the mean position of all the points in the object of interest.

thank you all am very greatful

There is no such a big difference between the avg coordinates of edge and of the segmented mask, I think. The circumscribed rectangle is used mostly, when detecting objects. But it was not the original question, of course :o)

They are not always same! Centroid obtained from edge image will be more biased to the smoothness of its boundary. Consider a circular shape where left half of its boundary is smooth and right half is rough. Lets consider the number of boundary pixels in right side is two times than that is in left. In this scenario, centroid from the edge image will be detected further right than it is detected from the segmented mask.

I didn't say that the results will be the same :o)

I just tried to note something about importance of (complex) calculation of centroid when, maybe, something simple is needed. I know, the original question was about centroid, OK :o)

Hi. How if I want to plot the circle on the same image cattle.jpg?