[Insight-users] Possible bug in itk::ShapeDetectionLevelSetFilter (or thereabouts)

Jahn Otto Næsgaard Andersen Jahn.O.Andersen@unimed.sintef.no
Wed, 25 Sep 2002 16:12:02 +0200


Hi,

I am having problems with a segmentation application that uses the 
itk::ShapeDetectionLevelSetFilter for processing of a 3-dimensional 
float image (as opposed to the level set test, which is using a 
2-dimensional image).

After some iterations, the itk::FastMarchingExtensionImageFilter (which 
is used by the ShapeDetectionLevelSetFilter) terminates with the error 
"itk::FastMarchingExtensionImageFilter(...) in Initialize(): 
AuxTrialValues is the wrong size.

What happens is that the number of trial values and trial points are 
not the same in the itk::LevelSetNeighbourhoodExtractor object (which 
is also used by the ShapeDetectionLevelSetFilter).

When the LevelSetNeighbourhoodExtractor::CalculateDistance(IndexType& 
index) function is called, somehow one of the values that are added to 
m_AuxOutsidevalues is skipped.

It seems like it might be a problem with the test that decides whether 
or not a value is an "outside value".

Btw: this worked with an earlier version of ITK. I am now using the 
current nightly version.


Best regards,
Jahn Otto