How can I find the centroids of closely seperated object?

In the left pannel, I have four obejct, they are close but separated. I am using the " regionprops( image ,'Area','BoundingBox', 'Centroid') but in return I am getting only one centriods. I need each object's centroids. Can you please help me with it?

[DGM]

Please attach a copy of the actual image you're using.

[Bipul Biswas]

Thanks for the message.

This is the original image.

Left pannel is the dark part of the partilces and right pannel is the bright part of the partilces (Top Figure). I need the centroids of both dark and black parts of each particles.

Answer 1

[DGM]

I'm going to go out on a limb here and guess that the image you're feeding to regionprops() is a numeric image instead of a logical image. If that's the case, it will assume that it's a label array (as from bwlabel()). From that assumption, it interprets all blobs as having the same label (label 1), treating them all as one object.

inpict = imread('4dot.png');

% if the image is numeric
inpict = im2double(inpict);
S = regionprops(inpict,'centroid');
C = vertcat(S.Centroid)

C = 1×2

204.0885 135.5302

% if the image is logical
inpict = inpict>0.5;
S = regionprops(inpict,'centroid');
C = vertcat(S.Centroid)

C = 4×2

166.3950 106.1014 184.9957 160.6012 232.1303 97.4901 262.4825 160.1700

[Bipul Biswas]

Thank you very Much.

Understood the logic and it's working.

[Image Analyst]

I would have added a call to bwconvhull on the segmented (binary) image to try to round out those particles after thresholding.

[DGM]

Given that OP is dividing the objects into two regions, I was under the impression that the half-object centroids was more important than the centroid of each round object.

[Bipul Biswas]

Yes, in this case, half part's centroid is more important.

BTW, Thanks.