[Insight-users] normalized histogram mutual information metric

[Luis Ibanez]

Hi Christos,

When you change the Metric of the Registration Framework,
you should readjust the learning rate of the optimizer
since its effect is proportional to the magnitude of the
metric gradient.

Whenever you get the Exception of non-overlap from the
registration framework, you should suspect of being using
a learning rate that is *too large*, and you should also
suspect of the translation scaling of the transforms.



    Regards,


       Luis



--------------------
Christos Panagiotou wrote:

> hi all
> 
> i am trying to use the normalized mutual information histogram image to 
> image metric.
> i ve altered the already existing insight application called 
> MultiResMIRegistration to integrate
> the above metric and an affine transformation instead of the existing 
> rigid body.
> 
> I wanted to integrate the normalized version of the metric as its 
> supposed to be invariant to overlap
> (Studholme et all, "An overlap invariant entropy measure of 3d medical 
> image alignment")
> 
> as i said i use the MultiResMIRegistration which centers the 
> images/volumes before the registration procedure.
> i ve the metric and an optimizer (gradient descent optimizer is what i 
> use at the moment)
> to the MultiResolutionImageRegistrationMethod however when i run it, i 
> still get the overlap exception ...
> 
> is this metric supposed to decrease the chances of the overlap exception 
> or it should totaly eliminate them and
> then something else wrong is happening with my code?
> 
> regards
> christos
> 
> _______________________________________________
> Insight-users mailing list
> Insight-users at itk.org
> http://www.itk.org/mailman/listinfo/insight-users
>