Geometric center of a group of particles?

[HoweLab]

Hi.

We routinely generate images comprising 'roughly' circular groups of small, circular or ellipsoid particles or blobs. There are typically 5000-10000 per group. Is there an easy (or easier) way to determine the geometric center of such a group - in other words, a way to get an X,Y coordinate that represents the center of the group? The blobs are generally similar in size, but can sometimes cluster or clump together. These clumps should not 'pull' the center position towards themselves - i.e. we're looking for geometric center, not center of mass (I apologize if I'm not using those terms correctly). Thanks, in advance, for your advice.

[Winnok H. De Vos]

Hi Alan

You could determine the maxima as a selection using find maxima (under  
process>binary) either on a binary image (if you already have a way of  
segmenting all the blobs) or on your original image (with a certain  
noise tolerance) and subsequently fit the convex hull (under  
edit>selection) after which you allow the analyze particles command to  
find the centroid. This is only an aproximating way but does not take  
the internal distribution nor size or intensity of the blobs into  
account, which seems to suit your needs.  Maybe this already helps.
Kind regards,

Winnok


On 09 Oct 2009, at 22:55, HoweLab wrote:

> Hi.
>
> We routinely generate images comprising 'roughly' circular groups of  
> small,
> circular or ellipsoid particles or blobs. There are typically  
> 5000-10000 per
> group. Is there an easy (or easier) way to determine the geometric  
> center of
> such a group - in other words, a way to get an X,Y coordinate that
> represents the center of the group? The blobs are generally similar  
> in size,
> but can sometimes cluster or clump together. These clumps should not  
> 'pull'
> the center position towards themselves - i.e. we're looking for  
> geometric
> center, not center of mass (I apologize if I'm not using those terms
> correctly). Thanks, in advance, for your advice.
> --
> View this message in context: http://n2.nabble.com/Geometric-center-of-a-group-of-particles-tp3797231p3797231.html
> Sent from the ImageJ mailing list archive at Nabble.com.

___________________________

Winnok H. De Vos, PhD

Bio-imaging and Cytometry Unit
Dept. Molecular Biotechnology
University of Ghent
Coupure Links 653
9000 Ghent, Belgium

Tel [phone]
Fax [phone]

[HoweLab]

That worked *perfectly*! Thanks very, very much Winnok.

-Alan

Winnok H. De Vos wrote

Hi Alan

You could determine the maxima as a selection using find maxima (under  
process>binary) either on a binary image (if you already have a way of  
segmenting all the blobs) or on your original image (with a certain  
noise tolerance) and subsequently fit the convex hull (under  
edit>selection) after which you allow the analyze particles command to  
find the centroid. This is only an aproximating way but does not take  
the internal distribution nor size or intensity of the blobs into  
account, which seems to suit your needs.  Maybe this already helps.
Kind regards,

Winnok


On 09 Oct 2009, at 22:55, HoweLab wrote:

> Hi.
>
> We routinely generate images comprising 'roughly' circular groups of  
> small,
> circular or ellipsoid particles or blobs. There are typically  
> 5000-10000 per
> group. Is there an easy (or easier) way to determine the geometric  
> center of
> such a group - in other words, a way to get an X,Y coordinate that
> represents the center of the group? The blobs are generally similar  
> in size,
> but can sometimes cluster or clump together. These clumps should not  
> 'pull'
> the center position towards themselves - i.e. we're looking for  
> geometric
> center, not center of mass (I apologize if I'm not using those terms
> correctly). Thanks, in advance, for your advice.
> --
> View this message in context: http://n2.nabble.com/Geometric-center-of-a-group-of-particles-tp3797231p3797231.html
> Sent from the ImageJ mailing list archive at Nabble.com.

___________________________

Winnok H. De Vos, PhD

Bio-imaging and Cytometry Unit
Dept. Molecular Biotechnology
University of Ghent
Coupure Links 653
9000 Ghent, Belgium

Tel [phone]
Fax [phone]

[Michael Doube]

Hi Alan,

the method I'm most familiar with is the centre of mass approach: you get the coordinates of all the foreground pixels and the mean of their x and y coordinates is the (x,y) coordinate of the centre of mass, which is the same as the geometric centre if the pixels are not weighted by their intensity (~pixel 'mass').

If you want the geometric centre of the group, perhaps the first thing to do is define the boundaries of the group, fill that boundary and find the centroid of the resulting shape?  Is this the sort of thing you want to do?  This makes the assumption that the group is homogenous, which is what I think you want when you say you want to avoid clumps affecting your result.

Mike
________________________________________
From: ImageJ Interest Group [[hidden email]] On Behalf Of HoweLab [[hidden email]]
Sent: 09 October 2009 21:55
To: [hidden email]
Subject: Geometric center of a group of particles?

Hi.

We routinely generate images comprising 'roughly' circular groups of small,
circular or ellipsoid particles or blobs. There are typically 5000-10000 per
group. Is there an easy (or easier) way to determine the geometric center of
such a group - in other words, a way to get an X,Y coordinate that
represents the center of the group? The blobs are generally similar in size,
but can sometimes cluster or clump together. These clumps should not 'pull'
the center position towards themselves - i.e. we're looking for geometric
center, not center of mass (I apologize if I'm not using those terms
correctly). Thanks, in advance, for your advice.
--
View this message in context: http://n2.nabble.com/Geometric-center-of-a-group-of-particles-tp3797231p3797231.html
Sent from the ImageJ mailing list archive at Nabble.com.