[Insight-users] FastSymmetricForcesDemonsRegistrationFunction

[Torsten Rohlfing]

Martin:

I don't remember when I last tried the SymmetricDemonsRegistrationFilter 
with the FastSymmetricDemonsRegistrationFunction. I am mostly using the 
"Fast" function with the curvature-PDE registration filter. In that 
context, there is a test in Testing/Code/Algorithms that covers this 
combination. Also, I am routinely using the curvature PDE registration 
with the FSDRFunction in two little command line tools. I have put the 
source for these up for you to download from

http://www.stanford.edu/~rohlfing/curv.tar.gz

Please give it a shot and see what the registration results look like. 
The tools are similar to the DemonsRegistration example that comes with ITK.

As for the interaction between the symmetric demons registration filter 
and the FSDRFunction: I have no quick answer, but if the metric values 
are going the wrong way, I would probably try to see if a quick hack 
like changing the sign of the gradient of the force function fixes the 
problem. I currently have no time to play with it, but you should easily 
be able to do that. If it works afterwards, let me know, so we can fix 
that consistently in the force function and the curvature registration 
filter.

TR