[Insight-developers] some edits to the registration parts of the why page

[Matt McCormick]

Looks good, Brian.

I had put together that page right before MICCAI, so feel free to
update as you see fit.  Please only add what has been merged.
Currently, I do not believe the evolutionary optimizers have been
ported or multi-component registration.

Re: multi-component gradients, what way did you decide to go with the
data-type: itk::Matrix?  A new a-symmetric covariant second rank
tensor?  Should we have a class for contravariant tensors?  A class
that can handle contravariant, covariant, mixed-variance?

Matt

On Fri, Dec 9, 2011 at 10:29 PM, brian avants <stnava at gmail.com> wrote:
> @  http://www.itk.org/Wiki/ITK_Release_4/Why_Switch_to_ITKv4
>
> current:
>
> New Registration Framework
> Automatic optimization parameter scale estimation
> Composite transform
> Displacement field transform
> Better integration of dense displacement field registration into the framework
> Better support for multi-threading
> Additional evolutionary optimizers
> Different B-Spline registration approach available
> Transform symmetric second rank tensors
>
> proposed:
>
> New unified and fully multi-threaded optimization and registration framework
> Unified framework supports rigid, affine, dense and composite
> registration  (e.g. Demons registration initialized by affine
> transform)
> Improved multi-threaded metrics for rigid, affine and deformable registration
> New metrics support sparse or dense sampling
> Metrics for landmark or label guided registration
> Automatic parameter scale estimation for registration
> Automatic step-size selection for gradient-based registration optimizers
> Composite transforms and composite transform optimization
> Displacement field and diffeomorphic velocity field-based transforms
> Better support for multi-threading in multiple components of
> registration and optimization hierarchy
> Additional evolutionary optimizers
> Improved B-Spline registration approach available
> Accurately transform and reorient covariant tensors and vectors
>
> Any other suggestions welcome,
>
> brian