[ITK-users] Segmentation fault when trying to generate a BSpline transform from a set of matching landmarks

[Carlos Henrique Villa Pinto]

Hi, everyone.

I have two sets of landmark points, detected from two different 3D MR
images, and I'm trying to estimate a nonlinear mapping between the pairs of
matching landmarks. I already have the matchings, I just need to generate a
BSpline transform from them. The LandmarkBasedTransformInitializer class
seems perfect for that, but when I call its InitializeTransform() method
(after setting all the required input parameters), the result is a
segmentation fault.

I'm using ITK 4.8.0. I checked the source code and made some tests, but the
only thing that I could find out is that the problem happens in the
GenerateData() method of the BSplineScatteredDataPointSetToImageFilter
class, when the SingleMethodExecute() method is called by the
ThreadGenerator object. In such version of ITK, that happens in line 314 of
the file BSplineScatteredDataPointSetToImageFilter.hxx. I couldn't go
further because I have no clue of what exactly is being called by the
ThreaderCallback in that class.

Does anyone have any clue about what sort of things could cause a
segmentation fault in the BSplineScatteredDataPointSetToImageFilter class?
Could that be a bug, maybe?

I don't know if this information is relevant, but my images have both LPI
orientation, origin = (-127.5, -127.5, -127.5) and spacing = 2 x 2 x 2 mm.
And I'm using the SetReferenceImage() method of the
LandmarkBasedTransformInitializer class to define the parametric domain of
the BSpline transform.

I can provide some pieces of code and more information about my data if
needed.

Thanks in advance!

[]s

-- 
Carlos Henrique Villa Pinto
Graduate Student in Computer Science
Federal University of São Carlos - Brazil
XCS
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://public.kitware.com/pipermail/insight-users/attachments/20151029/a513c5a2/attachment.html>