The following error occurs when a registration of two images is started with the "RigidRegistration" Bundle:
itk::ERROR: MattesMutualInformationImageToImageMetric(0x4008b90): Too many samples map outside moving image buffer: 5482 / 90000
also, MITK crashes with the following error message:
1193: matrix has non-finite elements
here it is:
-nan -nan -nan
nan nan nan
-nan -nan -nan
Seems to happen only if the image has no zero values at the border, for example pic3d.nrrd .
This is normally what happens when the first stepsize is to big or if in the initial situation too few points overlap. In this case the metric can't provide a good gradient for the optimizer. I think this has nothing to do with the absence of black borders. This is also no bug, but should be handled neatly
This error occurs without MITK really crashing. Ofcourse a messages that explains a bit clearer what could be the problem here would be nice, but that would be a feature request.
AFAIK, this is Jan's domain, so I am reassigning this bug. Jab, please forward the bug if you are not the appropriate assignee.
This will no longer be relevant once T20202: Deprecate our registration in favor of MatchPoint has been resolved.