[Insight-developers] Notes from TCON of May 25, 2001

[Will Schroeder]

Hi Folks-

Here are my notes from today's telephone conference. If you have any questions or concerns please get back to the list. Some of these action items were due to a recent refactoring here at Kitware with GE, which made up much of the discussion today.

+ Jim Miller/ Dan Blezek of GE reworked the testing dashboard to make it more modular. The goal is to make it pluggable into any software system, especially ITK applications, etc.

+ Jim Miller added a project (Insight/Examples/SampleProject) that can be built outside of Insight's source tree using CMake. (Check the mailing list.) He's checking on the Unix side of things. This i in response to George's request.

+ Discussion item: Definition of a voxel/Voronoi tile: Celina agreed to continue the discussion via email.

+ DefomableMesh to be split into a process object (SphereSource) that outputs itkMesh (will will do it)

+ Voronoi2DDiagram ditto

+ AffineTransformation -> renamed to AffineTransform (will)

+ itkTransformation -> renamed itkTransform to be consistent with other classnames (will)

+ AffineTransformMeshFilter -> renamed itkTransformMeshFilter (will)

+ Discussion item: Nesting namespaces: there are many "implementation" classes that show up in the itk:: namespace. This clutters the documentation, etc. The thinking is to continue what Luis has started and create several categories of namespace. For example a itk::Diffusion:: namespace. There was some talk about creating another layer, such as itk::Diffusion::Details (similar to what the C++ boost folks do). This needs thought and discussion by all.

+ Reminder: please use ProgressMethods in your filters.

+ CentralDerivativeImageFilter -> renamed to CentralDifferenceImageFilter (will)

+ Reminder: there are a lot of bad method names and violations of the style. Example: SetNumClasses. Please use the long names and learn how to use Emacs completion.

+ Discussion item: Classifier vs. Labeler. Are these the same thing? Should the current classifier be converted to a filter? or is it something different?

+ Discussion item: Should a histogram BE a image or USE a image. Also, the naming was discussed: StatDenseHistogram. This relates to namespaces. Maybe itk::Statistics::DenseHistogram? And what should the file be named? itkStatDenseHistogram or itkDenseHistogram?

+ DynamicBorderArayKLM -> KLMDynamicBorder (will)

+ DynamicPolygonCell -> PolygonCell (will)

+ DynamicPolygonBoundary -> PolygonBoundary (will)

+ Discussion item: Use of NumericTraits for proper computation precision. Example: AvgGradMagSquare in itkAnisotropicDiffusionImageFilter. It looks like the precision of the accumulator and counter is bad, since they are related to the TPixelType template parameter. We need to think through this: how to make sure that the precision is high enough; how to make sure that the output image is of high-enough precision.

+ Luis has been adding groups to doxygen for the purposes of some high-level, on-line documentation. See http://public.kitware.com/Insight/Doxygen/html/modules.html. Suggestions for some other modules were made.

+ CMake is being packages into a pre-compiled, zipped, self-installing package. Windows user will be able to install very easily. Will keep the list posted.

Will