[Insight-users] ImageMomentsCalculator question

[Benjamin King]

Hi Luis,

>             "Principal moments"
>                   *ARE*
>         "variances of the sample
>          along the principal axes".

Here's a snippet of the output my test produces:

-----
Moments after principal axis transformation
[...]
Second central moments:
1147.43 -0.595038
-0.595038 6623.77

Principal Moments: [1.0136e+10, 5.85124e+10]
-----

The principal moments are not the eigenvalues of the second central moments 
matrix, but from your explanation I understand that the should be.

I found the reason for this in line 172 of itkImageMomentsCalculator.txx:
The eigenvalues of the second central moments matrix are multiplied with the 
total mass to give the vector reported by GetPrincipalMoments().

Is that a feature? To me, that behaviour wasn't obvious from the documentation 
or the comments in the code, so it feels more like a bug.

I did not find any usage of GetPrincipalMoments() in neither the ITK source 
nor the Insight Applications, so I think it would not break much code if the 
multiplication would be removed.


Cheers,
	Benjamin