You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
As @dzenanz pointed in #20 , the module needs to be refactored. Some classes seem to be specialized versions of classes that exist in the toolkit (Canny detection filters, morphological filters, etc.), but it should be investigated whether better inheritance can be achieved.
Some minor issues related to the refactoring, or which could be done while refactoring are:
Offset-table copy sized by sizeof(unsigned long) in VotingBinaryHoleFillFloodingImageFilter and RegionCompetitionImageFilter fixed; it left the slice and volume strides uninitialized and crashed on Windows with ITK 6 (COMP: Build and test against ITK 5.4 and ITK 6 #61).
Still open from the original list, in rough priority:
Port CannyEdgeDetectionRecursiveGaussianImageFilter from ITK_THREAD_RETURN_TYPE callbacks and ProgressReporter to DynamicThreadedGenerateData and ProgressTransformer (moved here from Compiler error concerning ITK_THREAD_RETURN_VALUE #20).
Decide whether that filter, DescoteauxSheetnessImageFilter, and the morphology helpers duplicate toolkit classes closely enough to be replaced.
Header/implementation split, Set/Get macros, default member values, and complete PrintSelf across the generators and aggregators.
Items 1 and 2 of the 2026-10-08 status (the Canny filter threading port and its duplication of the ITK filter) moved to #63, which tracks bringing CannyEdgeDetectionRecursiveGaussianImageFilter up to the fixes in itk::CannyEdgeDetectionImageFilter. The remaining items here (header/implementation split, Set/Get macros, default values, PrintSelf, and the Descoteaux and morphology helpers) stay in this issue.
Activity
jhlegarreta commentedon Oct 11, 2018
Hopefully, these two branches, when they eventually get more work and get merged, may help in this tremendous task:
hjmjohnson commentedon Oct 8, 2026
Status as of 2026-10-08, so this stays a live tracking issue rather than a stale wish list.
Done or in review:
ITK_EXERCISE_BASIC_OBJECT_METHODSsuperclass names corrected, BUG: Name the real superclass in ITK_EXERCISE_BASIC_OBJECT_METHODS calls #59.include(${ITK_USE_FILE})removed from the external build so the test driver links, COMP: Do not include UseITK when building as an external module #60.sizeof(unsigned long)inVotingBinaryHoleFillFloodingImageFilterandRegionCompetitionImageFilterfixed; it left the slice and volume strides uninitialized and crashed on Windows with ITK 6 (COMP: Build and test against ITK 5.4 and ITK 6 #61).Still open from the original list, in rough priority:
CannyEdgeDetectionRecursiveGaussianImageFilterfromITK_THREAD_RETURN_TYPEcallbacks andProgressReportertoDynamicThreadedGenerateDataandProgressTransformer(moved here from Compiler error concerning ITK_THREAD_RETURN_VALUE #20).DescoteauxSheetnessImageFilter, and the morphology helpers duplicate toolkit classes closely enough to be replaced.Set/Getmacros, default member values, and completePrintSelfacross the generators and aggregators.Examples/andUtilities/Visualizationstill call the pre-ITK5SpatialObjectPoint::GetPosition()and only build withLSTK_USE_VTK=ON, which CI never exercises (Unable to successfully compile LesionSizingToolkit module #42).hjmjohnson commentedon Oct 13, 2026
Items 1 and 2 of the 2026-10-08 status (the Canny filter threading port and its duplication of the ITK filter) moved to #63, which tracks bringing
CannyEdgeDetectionRecursiveGaussianImageFilterup to the fixes initk::CannyEdgeDetectionImageFilter. The remaining items here (header/implementation split,Set/Getmacros, default values,PrintSelf, and the Descoteaux and morphology helpers) stay in this issue.