[Insight-users] smoothing by AntiAliasBinaryImageFilter, Gaussian blur or windowedSincFilter

[lynx.abraxas at freenet.de]

Hello!


Let's  say  I  have  a  Platonic  solid  as a mesh (or even simpler a plane in
[1,1,0] or [1,1,1] direction) I voxelize that to get a 3d-binary image. Then I
wonder  what  would  be  the  best way to get back to a mesh who's shape is as
close as possible to the original mesh.
Just using marching cubes introduces uneven surfaces. From the vtk users  guid
I  know gaussian smoothing the binary image can create a much smoother result.
I was told AntiAliasBinaryImageFilter does something similar and then there is
also the possibility to use some mesh smooting filter like Laplacian smoothing
(vtkSmoothPolyDataFilter, used in vtk-book) and vtkWindowedSincPolyDataFilter.
So  there  are two different classes of filter 1. voxel filter 2. mesh filter.
Now I wonder:

- Which one is the best for my problem?
- Would it make sense to use more than one filter?
- How can the voxel filter result in a  smoother  mesh  without  the  need  to
subsample  the  voxel data? Or asked differently: How can the mesh be smoother
altough the voxel discretisation stays the same???
- What is the difference between Laplacian and Sinc smoothing or what type  is
better?
-  What  is  the difference between Gaussian and AntiAliasBinaryImageFilter or
which one is better?
- What filter is least dependant on abritrary chosen parameters?


Sorry for the many questions but I'd be very thankful for any answers.
Lynx