[Insight-users] Re: AffineTransform

[Luis Ibanez]

Hi Silvester

The center of rotation in the default AffineTransform is
assumed to be in the origin of coordinates.

This is a very sensitive issue since in general ITK images
have their origin in one of the image corners as opposed
to the image center.

Recently we have experimented with a variant of this
Transform, the :  itk::CenteredAffineTransform

http://www.itk.org/Insight/Doxygen/html/classitk_1_1CenteredAffineTransform.html

In this transform you can define where the center of rotation
is. An example on the use of this transform for performing
registration is available in Section 7.5.3, pdf-page 199.

The associated code is available at

   Insight/Examples/Registration/ImageRegistration9.cxx

The Centered transforms are a topic in progress and discussion is
going on about the best way in which they should be implemented.

The open project for this topic is available at:

http://caddlab.rad.unc.edu/itkProjects/

more specifically

http://caddlab.rad.unc.edu/itkProjects/indexproject.php?id=8

You are wellcome to join forum on this project and contribute
to the open discussion.

For more on the notion of Image origin, you may want to look
at the section on the itk::ResampleImageFilter. Section 5.7.1
pdf-page 139.


Please let us know if you have further questions,


Thanks


    Luis


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

Silvester Czanner wrote:
> Luis,
> 
> In the Software Guide there in the chapter 7.7.12 is written, that the 
> itk::AffineTransform represents an affine transformation composed from 
> rotation, scaling and translation. I would like to ask you, how is here 
> defined the center of rotation. Thank you.
> 
> Silvester
> 
> -- 
> 
> Silvester Czanner, PhD
> 
> Pittsburgh Supercomputing Center    | tel. +1 412 268 3496
> Carnegie Mellon University          | fax  +1 412 268 5832
> 4400 Fifth Avenue,                  | czanner@psc.edu
> Pittsburgh, PA 15213, USA           | http://www.psc.edu/~czanner
> 
> ... the life would be much simpler, if we have a source code for it ...
> 
> 
> 
> 
> 
> 
>