[Insight-developers] Adopting QE for simplex (2)

[Alexandre GOUAILLARD]

Hi guys,

Here are some quick answers:
You are suppose to switch back and forth from itkQEMesh and itkMesh without
problem. Just change MeshType. Templates are the same, in the same order to
make this easier.

Then, for the simplex mesh case, the good thing is, dual mesh connectivity
is already there. So the triangleToSimplex and SimplexToTriangle classes do
not have any reason to be anymore. Unfortunatly, I did not come with a
second container with dual geometry (points). That's where I wanted to work
a little bit for a new simplex solution. I have a working prototype for that
though, which can use a functor to define the geometry (barycenter of
triangle by default). I think you would need to adapt some filters, as the
force filters expect itkMesh API.

All eulers operators have been extensively tested (see
\Testing\Code\Review). But it looks like there is a few bug remaining,
especially in flipedge.

I did not test with deformable model. But I'm definitly in! Keep some of the
fun for me :-D

Don't hesitate to send in more questions. I should be online once a day
here. Send me in a little piece of code with itk::Mesh (like, a minimum
pipeline for simplex) and I should make it work with itkQE in a few hour (or
not :) ).

Note, there is an initialisation cost. If the simplex mesh is static (no
topological changes) then, it might not be worth it, in term of speed. If
you intend to modify the mesh (subdivide faces where they become too
elongated) then it's **definitly** worth it. No need for the internal cell
tables and everything you coded within the simplexmesh structure to bypass
BuildLInk().

Alex.