[Insight-users] MI rigid registration

[Luis Ibanez]

Hi John,

It seems to be a good practice to normalize the intensities of your
images before using them for registration with MI.

This is a statistical normalization, not a range one. So the goal
is to obtain an image with mean intensity value equal to zero, and
standard deviation of intensities equal to 1.0.

You may use the itk::NormalizeImageFilter for this purpose.
http://www . itk . org/Insight/Doxygen/html/classitk_1_1NormalizeImageFilter . html


Then, start your registration with *very*  small values of the step 
length and try using a simple transform first (e.g. just translation)
until you get a better feeling on the parameters appropriate for your
images.

Plotting the trace of the metric values, and the paramters of the
transform as the registration progresses is in general a great help
for getting insight in how to tune the parameters.

You may want to take a closer look at the registration chapter that
Lydia Ng setup in the SoftwareGuide:

             http://www . itk . org/ItkSoftwareGuide . pdf

Section 7.4, pdf-page 218 to 225.

You will find, for example, the use of normalization filters there.



Regards,


    Luis


------------------------------------

Dill, John wrote:
> I have been having problems trying to find parameters to use MI registration
> on DRR and portfilm images with MutualInformationEuler2DRegistration.  I was
> wondering if there were any hints on how to find parameters.  The
> combinations that I've tried do not seem to lock onto anything but continue
> to float farther and farther away as more iterations are used.  I am trying
> to register a DRR image and a portfilm, which is similar to DRR but with
> less contrast.  Are there any intensity distribution considerations that I
> am not aware of?
> 
> If anyone is willing to take a closer look, I can send images to an email.
> 
> Thanks for any help.
> John
> _______________________________________________
> Insight-users mailing list
> Insight-users at itk . org
> http://www . itk . org/mailman/listinfo/insight-users
>