[Insight-users] ConnectedComponentImageFilter and NumberOfThreads bug.

Greg Harris Gregory-Harris at uiowa.edu
Mon Sep 22 13:32:05 EDT 2008

ITK Folks,

Since the changes of 31 July 2008 to itk::ConnectedComponentImageFilter that
added multi-threading support, we have observed that when we give it a
3-dimensional binary brain mask image, we get different results if we leave
the number of threads greater than 1.  The principal component is split half
way through the image in the z direction.  The workaround is to do a
SetNumberOfThreads(1) before the Update().

Someone will probably want to add a 3-dimensional test case to the testing
suite.  I could correspond with them about a nifti format brain mask image
file that would exhibit this behavior, if needed.

Greg Harris
Iowa Psychiatry Brain Imaging Lab

More information about the Insight-users mailing list