[Insight-users] Registration using itk::OrientedImage

[Rupert Brooks]

Sorry about that, i actually did try to search the bug database, but
missed it somehow.

> There are a couple of suggested solutions...
>
> The challenge here is how to fix the computation for
> the case of the OrientedImage without penalizing
> the computation time for the standard itk::Image.

Or without requiring every single metric that computes a gradient to
remember to fix this.  That could be a lot of duplicated code.

>
>
> Suggestions are welcome.

I ~have~ been thinking about it actually :-)
Another efficiency issue / opportunity is that images that have zero
origin, or unit spacing can be calculated more efficiently.  Are you
dealing with that in the recently proposed efficient image
registration framework?  I haven't had time to plow through all the
code yet.

In any case, i cant think of a clean way around the problem besides
refactoring to have itk::Image be a subclass of itk::OrientedImage.
I've worked on it a bit and theres also issues related to not
templating virtual functions.

I should probably formalize my thoughts in Insight Journal form rather
than rambling on about it here.  If anyone else is working on that,
I'm happy to collaborate.

Cheers,
Rupert B.

-- 
--------------------------------------------------------------
Rupert Brooks
McGill Centre for Intelligent Machines (www.cim.mcgill.ca)
Ph.D Student, Electrical and Computer Engineering
http://www.cyberus.ca/~rbrooks