[Insight-developers] Finite Difference denoting filter failures

[Bradley Lowekamp]

Hello,

On or about September 20th it appears that ITK some how changed the results of the DenseFiniteDifference denoting filters:

http://www.cdash.org/CDash/viewTest.php?onlyfailed&buildid=1568445

This particular SimpleITK build is build of SimpleITK's stable master ( which hadn't changed since the 16th when the failure started ), against the nightly ITKv4.

Simple ITK has baseline images for the BinaryMinMaxCurvatureFlow, CurvatureAnisotropicDiffusion, CurvatureFlow and Gradient AnisotropicDiffusion image filters. They are no longer passing the provided tolerances. The computed images changed.

The only thing I see that changed in ITKv4 that day was the addition of the named Data Object Inputs. I however don't see how that relates.

Anyone have any ideas what's the cause in this change of behavior?

Thanks,
Brad