[Insight-developers] OrientedImage - Direction Cosines - Gradients - ImageRegistration

[Luis Ibanez]

As you probably know, currently we are making an inconsistent
use of the image orientation in the image registration framework.

Due to its difficulty, we postponed addressing this issue until
after we have cut the release of ITK 3.4.

So, here we are, eager to fix the problem now.

The central point in this issue is the fact that Image Metrics
compute image gradients as part of the process of computing
their metric derivatives.  The helper classes that compute
image gradients, however, do not take the image orientation
into account.

The result is that image metric derivatives are computed
incorrectly when itk::OrientedImages are used.

The issue is reported as Bug 5081:
http://public.kitware.com/Bug/view.php?id=5081
(and probably in a second bug report too...)


We have created a proposal page to craft a solution to this
issue. You will find this page in the Wiki:

   http://www.itk.org/Wiki/ITK_Oversight_Committee#Proposals

more specifically at:

   http://www.itk.org/Wiki/Proposals:Oriented_Image_Registration


At this point we are trying to define the naming of the new
methods to be added to the itk::Image and itk::OrientedImage
classes. This naming is critical due to the high visibility
of these classes, so we want to think hard on these names to
make sure that we don't find compelling reasons to change them
later.


Please add your comments to the Wiki page.


     Thanks


        Luis