[Insight-users] 3D rigid image registration

[Stephen R. Aylward]

Hi,

We have also created a more-full-featured registration application using 
itk.   It will soon be made part of ITK, but until then, you can 
download source and binaries from
http://caddlab.rad.unc.edu/software/index.html#registration3D

You begin by placing four landmarks in each image, the pre-registration 
step performs rigid squared-error minimization to align those landmarks, 
and then the registration process refines that registration using mutual 
information registration.   All is within the ITK framework and the 
source is provided. This project will be in the ITK CVS repository after 
the first of the year.

However, to learn how to do it yourself, the application you found is 
probably a better starting point....

In regard to question (1), mutual information is good for inter and 
intra-modality registrations.   Other methods may be better or worse, 
but mutual information registration has been demonstrated to be good.

I don't have the details on the parameter file, but I sure you'll get 
another email on that soon...

Stephen


Steve Boyd wrote:
> Hello,
> 
> I use micro-computed tomography to generate 3D image data of bone
> micro-structure, and I'd like to register images taken at different time
> points. It seems that the example in the ITK kit for mutual information
> registration would be a suitable approach to determine the 3D rigid
> transformation. I have two questions: 
> 
> 1. Because my data is not multi-modal (all from CT), is it a
> disadvantage to use this approach?  Is there another approach that I
> should consider?
> 
> 2. Trying to run the ITK example "MultiResMIRegistration" I need to
> include a parameter file as an argument, but I do not know what this
> file should contain.  Have I missed downloading a part of the ITK
> package, or can somebody explain what the param.file should contain?
> 
> Thank you for any help with either of these two questions.
> 
> Sincerely,
> 
> Steve
> 


-- 
===============================================
Dr. Stephen R. Aylward
Assistant Professor of Radiology
Adjunct Assistant Professor of Computer Science
http://caddlab.rad.unc.edu
aylward@unc.edu
(919) 966-9695