[Insight-developers] contrib: fast curvature PDE registration

[Torsten Rohlfing]

Luis --

You are welcome :))

Unfortunately, I found a problem with one of the included classes, the 
fast symmetric demons force function. Turns out (my fault) that it 
didn't work properly in the multi-resolution framework. Now that I 
identified and fixed the problem, I am not sure how it actually worked 
in the single-resolution environment. Anyway, turns out I forgot to set 
image origin and spacing in the local warp image filter. Attached is a 
tar archive with replacements for 
itkFastSymmetricForcesDemonsRegistrationFunction.{h,txx}.

Note that I also changed the following - the mean squared difference 
reported by ComputeUpdate() is now NO LONGER with respect to the updated 
transformation (see my earlier posting regarding 
itk::SymmetricForcesDemonsRegistrationFunction). Instead, the MSD is 
computed for the deformation field as it is upon entering the current 
iteration. I doubt this breaks anything, but it means that for now the 
fast implementation and the original class will report different values 
here.

On the upside, this saves some CPU time since we don't have to compute 
the new mapping of each pixel. Also, I find it much more consistent 
since the actual applied update is not necessarily identical to the one 
computed here anyway.

Best,
  Torsten

-- 
Torsten Rohlfing, PhD          SRI International, Neuroscience Program
 Research Scientist             333 Ravenswood Ave, Menlo Park, CA 94025
  Phone: ++1 (650) 859-3379      Fax: ++1 (650) 859-2743
   torsten at synapse.sri.com        http://www.stanford.edu/~rohlfing/

     "Though this be madness, yet there is a method in't"

-------------- next part --------------
A non-text attachment was scrubbed...
Name: fix.tar.gz
Type: application/x-gzip
Size: 4739 bytes
Desc: not available
Url : http://www.itk.org/mailman/private/insight-developers/attachments/20050505/b1fcfed9/fix.tar.bin