[Insight-users] gaussian weighting in MultiResMIreg.

[Luis Ibanez]

Hi Nasser,

Altering the intensities of a region (ROI) in the input
images will not make the MutualInformation metric to focus
(give more weight) in that region. Mutual information matches
the simultaneous occurrence of two intensity values regardless
of the actual value of the intensities. In other words, as far
as MI is concerned, image intensities are "labels" rathen than
scalar values.

Your approach may have worked in a MeanSquare metric and probably
in a NormalizedCorrelation metric.

If your final goal is to make the metric give more weight to
a particular region of the image, then you may want to join
the discussion about non-data pixels, and masking regions in
image metric.
http://www . itk . org/pipermail/insight-users/2003-October/005159 . html

In this approach, the metric will take as extra arguments two
masks that will define what pixels should be taking into account
for the computation of the metric.


Regards,


   Luis


--------------
Nasser Kashou wrote:
> Hello,
> Have a question about any preprocessing that's done on the images in the multiresmiregistration example.   Currently, I'm applying a spherical gaussian(std=5,u=0, samples=30, where it is 1 at infinity rather than 0) to a specific ROI in one of my datasets, before running it on the ITK code.  My question is how would the default gaussian weighting to estimate the entropies affect the preprocessing step that I made?  My purpose is to stress/weight the ROI that I chose when running MI.  Do I have the right approach or concept? Or does running the parzen window estimates nullify the weighting that I did outside of ITK?
> 
> Thank You
> 
> 
> _______________________________________________
> Insight-users mailing list
> Insight-users at itk . org
> http://www . itk . org/mailman/listinfo/insight-users
>