CGAL users discussion list

[Stephane Tayeb <>]

Ianic wrote:
> Stephane,
>
> That is easier said than done for someone like me, who is just making a
> discreet foray into CGAL workings... My naive thinking was that I output
> surface mesh as .off file and then read it in as in the elephant example.
> But I suppose you are suggesting something different here. Could you
> elaborate just a tiny bit more on your approach?

Hi,

If you output a surface mesh as .off file and then read it, you may 
encounter the same problems as before, because you will have no 
guarantee that the output is manifold.

Something which might work on your input is to reduce the approximation 
error and/or the size parameter of the surface facets in order to be 
sure to separate the surfaces (I guess the non-manifoldness of your 
output is related to this).


What I suggest in my previous answer is to use the surface mesher on 
your input with the manifold tag (be careful, if your input is not 
manifold, then the process will never end).

You will get a c2t3. You can insert the vertices of the triangulation of 
this c2t3 into a c3t3, then call refine_mesh_3 on this c3t3.

The technical point is to you use the same oracle for the surface mesh 
and the 3D mesh generation. The simplest way is probably to implement an 
implicit function for your voxelized image (which returns -1 inside, 1 
outside, see 'include/CGAL/Mesh_3/Image_to_labeled_function_wrapper.h' 
for instance) and plug it in implicit oracles of meshers.

Is it detailed enough to get a start ?

Regards,
Stéphane.

--
Stephane Tayeb
Software engineer - INRIA Sophia Antipolis
Geometrica Project-Team