[Insight-users] MeanSquaresImageToImageMetric: Too many samples map outside moving image buffer

[Urlek]

Hi,
   I am trying to register two considerably (but purposely) offset images
that share only approx. 30% of the surface. Somehow,
MeanSquaresImageToImageMetric throws the following exception:

itk::ExceptionObject (000000000113CC88)
Location: "void __cdecl itk::MeanSquaresImageToImageMetric<class
itk::Image<float,3>,class itk::Image<float,3> >::GetValueAndDerivative(const
class itk::Array<double> &,double &,class itk::Array<double> &) const" 
File:
e:\itk\insighttoolkit-3.8.0-src\code\review\itkOptMeanSquaresImageToImageMetric.txx
Line: 284
Description: itk::ERROR: MeanSquaresImageToImageMetric(0000000030CE1CE0):
Too many samples map outside moving image buffer: 219300 / 909500

This makes me wonder about the following questions:
1) Does that mean I have to cut out the regions that are likely to overlap
prior to the registration? 
2) If so, is it possible to define a region of the image to be used in
registration without having to cut that region out into a smaller image?
3) Would some metric other than MeanSquaresImageToImageMetric work better?
Preferably another crosscorrelation-like one that supports multi-threading? 

Thanks a lot!

-- 
View this message in context: http://www.nabble.com/MeanSquaresImageToImageMetric%3A-Too-many-samples-map-outside-moving-image-buffer-tp20660013p20660013.html
Sent from the ITK - Users mailing list archive at Nabble.com.