[vtkusers] Filling a hole

[Kevin H. Hobbs]

On Sat, 2006-04-22 at 01:02 +0800, Alexandre GOUAILLARD wrote:
> I did not try so flame me up alive if you want, 

Never! I'm learning a lot here. You posted the code that allowed me to
do that learning, and to pretty up the clipped edges of my data. There
will be no flaming from me. 

> but I think that if you 
> subdivide and decimate the patch itself , without taking care of its 
> original boundary, then it does not fit the original hole border you had 
> in the first place, right?

I thought, that vtkDecimatePro::BoundaryVertexDeletionOff() would leave
the edge of the patch be.  I see that is not exactly the case. The patch
_looks_ very good but it was off by enough that vtkCleanPolyData did not
connect it. A very small tolerance took care of that.

> Then when you try to patch up your mesh with the cap you just computed 
> you either have T-vertices (impossible with vtk though) or ... new holes.

Yup, now that you make me think about it, the subdivision might break
the cells along the boundary. It, might be responsible for moving the
edge of the patch. I have not read the papers, but I'm doing little
tests. What I do not see is anything that lets me explicitly tell it not
to break some cells.

> Note that you might actually "see" it if the boudaries (geometricaly) 
> match each other, but a call to feature edge should report new border edges.

I bet you're right, and if the topology isn't just right these new edges
might not be loops so we couldn't patch them. At least it looks lovely. 

> If you subdivide and then decimate globally, you don t have that 
> problem, but you don t really wanna do that though, right ?
> 

No, not really... 

Anyway, thanks for discussing this with me, it's been really
informative.

-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: This is a digitally signed message part
URL: <http://www.vtk.org/pipermail/vtkusers/attachments/20060421/c137a047/attachment.pgp>