[Insight-developers] contrib: fast curvature PDE registration

[Torsten Rohlfing]

Greetings --

I implemented a new PDE registration algorithm based on a paper by B. 
Fischer and J. Modersitzki, "A unified approach to fast image 
registration and a new curvature based registration technique," Linear 
Algebra and its Applications,  vol. 380, pp. 107-124, 2004. The 
implementation supports both 2D and 3D images.

The algorithm uses fast DCT (implemented using FFTW, which is necessary 
to use this algorithm) to perform registration using a curvature-based 
regularization in time O(n log n) per iteration, where n is the number 
of image pixels. The implementation fits into the PDE registration 
framework in ITK and can, for example, be used in a multiresolution 
fashion (see included example). Also, the registration filter class is 
templated over the image force function and can fundamentally be used 
with any existing registration function (e.g., mean squares, demons, 
etc.). A faster implementation of the symmetric demons forces function 
using a cached deformed moving image for faster gradient computation is 
also included.

There is a test for the curvature registration filter using the fast 
symmetric demons forces included, which is derived from the existing 
test of the demons algorithm.

The sources are available at

  http://www.stanford.edu/~rohlfing/itk_curvreg.tgz

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"