[Insight-users] inconsistency in MeanSquaredImageToImageMetric

[Michael Jackson]

I have been developing some code over the weekend that basically  
rotates an image about a fixed point, extracts a section of the image  
(ROIImageFilter) and compares that extracted image to another image  
from a "Fixed" image. Running the program I will get 1 of several  
different results:

1: Segfault
2: Runs, but the values are not correct
3: Runs, but the value _are_ correct
4: Partly runs but will throw an exeception after any number of  
iterations:

Angle	-0.4	Metric	6934.48
Angle	-0.38	Metric	6653.89
terminate called after throwing an instance of 'itk::ExceptionObject'
   what():  /Users/Shared/OpenSource/InsightToolkit-3.14.0/Code/Review/ 
itkOptMeanSquaresImageToImageMetric.txx:162:
itk::ERROR: MeanSquaresImageToImageMetric(0x2266bb0): Too many samples  
map outside moving image buffer: 0 / 640000

This is with itk 3.14.0 freshly compiled with Dynamic libs on OS X  
10.5.7 intel. This also happens with itk 3.10.1 on the same system  
with Static libraries.

  A few questions: Is there a "random" element to the  
MeanSquaredImageToImageMetric? Has anyone else ever seen behavior like  
this before?

  I have the a-priori knowledge of what the "correct" answer should be  
(or at least pretty close to what it should be) but I am working to  
make this an automated system.

Thanks for any help or explanations.
_________________________________________________________
Mike Jackson                  mike.jackson at bluequartz.net
BlueQuartz Software                    www.bluequartz.net
Principal Software Engineer                  Dayton, Ohio