[Insight-users] Questions about the Diffeomorphic Demons Registration

[Chunfang Wang]

Hello All,

I have a problem with the Diffeomorphic Demons registration.
Hope anyone can give me some suggestions.
Thanks in advance.

I used the DiffeomorphicDemonsRegistrationWithMaskFilter with two level
resolutions,
the main parameters of the filter is below :

MaximumRMSError = 0.01;
MaximalStepLength[ ] = {2, 1};
DemonsNumOfIterations[ ] =  {6, 10};
double DemonsElasticSigma [ ] = {0.8, 0.4} ;
double DemonsViscousSigma[ ] = {0.8, 0.4} ;

I set an observer of the filter to print out the RMS change  in each
iteration.
The registration results is quite well if the registration is run
sucessfully.
However, the result is not stable. In many cases, the observer print out the
RMS change is "nan",  and no deformation field is generated.
The program can run through in this situation, and there is no exception
error caught .

The normal RMS change of my test images always goes from around 0.3 to 0.1,
so I think it is not the reason that the value is nonreprentable, which is
the defination of "nan".
The same images with the same parameters in the program can obtain sucessful
registration results sometime, and failed with the RMS changes "nan".

You need to run the program four or five times, and possible the
registration will be sucessful once  if i am lucky enough.

I don't know what the reason of this unstable problem,  and hope someone can
figure me out.

Thanks.

Kind regards,
Chunfang
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.itk.org/pipermail/insight-users/attachments/20110512/94ff2cd3/attachment.htm>