Micheal, The only contribution i can offer is like this : i think the centroid function is in the Image processing ToolBox, but you can proceed as the following :
Suppose your Image I is RGB ( or HSV) you convert it to grayscale :
X=(0.2989*I(:,:,1)+0.5870 *I(:,:,2)+0.1140*I(:,:,3));
You make a binary sample by choosing a certain threshold, lets say the mean of the whole sample :
thresh=mean(X(:));
[m n]=size(X);
B=zeros(m,n);
for x=1:m
for y=1:n
if X(x,y)>thresh
G(x,y)=1;
end
end
end
Next you choose a tiny portion, let say a elementary particle from your sample ( in the medium as example ):
S=G(floor(end/2):10+floor(end/2),floor(end/2):10+floor(end/2));
You try next to count the white pixels, and try to use some geometrical relation to deduce the Radius,
