CGAL users discussion list

[Graham Macpherson <>]

Mariette, Pierre,

Thanks, that was it.

I had a feeling that there was tolerance adjustment somewhere that I had 
missed.

Interestingly, the tighter tolerance makes the mesher run faster for the 
sphere, to produce essentially the same number of points:

1e-3 - 13.6s
1e-5 - 10.2s
1e-7 - 10.5s

All the best,

Graham


On Tuesday 18 August 2009 08:13:18 Mariette Yvinec wrote:
> I think that there is some mistake.
> You are not talking about the same parameter.
> Pierre was  obviously talking about the third parameter
> in the  creator of the Implicit_surface_3
> and not about the third meshing criteria ....
>
> There is a  third parameter in this creator
>  which sets the precision
> with which points on the surface are computed.
> Implicit_surface_3<Traits, Function> surface ( Function f, Sphere_3
> bounding_sphere, FT error_bound = FT(1e-3));
> This parameter is relative to the radius
> of the bounding sphere and has a default value of 1e-03.
> Implicit_surface_3<Traits, Function> surface ( Function f, Sphere_3
> bounding_sphere, FT error_bound = FT(1e-3));
>  You need to take care that  the resulting error bound
> (i. e. error_bound times radius of bounding sphere) is no more
> than about 1/10 times the distance bound meshing criteria,
> and this might be the source of yoour troubles.
>
> Graham Macpherson wrote:
> > Hello Pierre,
> >
> > That parameter was at 0.001 in my initial post, if I change it to 0.0008
> > it triples the number of points on the surface (they form in clumps, not
> > uniformly distributed), but there is exactly the same magnitude of radius
> > error.
> >
> > If I make it 0.0005, my PC runs out of memory (8GB RAM) before it
> > completes.
> >
> > If I make it 0.1, I get essentially the same result as 0.001.
> >
> > Regards,
> >
> > Graham
> >
> > On Friday 14 August 2009 14:27:32 Pierre Alliez wrote:
> >> hi Graham,
> >>
> >> have your tried tuning the third parameter if the surface? the relative
> >> precision of the dichotomy mechanism to find the intersection?
> >>
> >> Pierre Alliez
> >> INRIA Sophia Antipolis - Mediterranee
> >> Project-team GEOMETRICA
> >> http://www-sop.inria.fr/members/Pierre.Alliez/
> >> Tel: +33 4 92 38 76 77
> >> Fax: +33 4 97 15 53 95
> >>
> >> Graham Macpherson a écrit :
> >>> Hello,
> >>>
> >>> I'm meshing surfaces using the implicit surface mesher and testing it
> >>> with simple smooth shapes.  Using the example
> >>>
> >>>     examples/Surface_mesher/mesh_an_implicit_function.cpp
> >>>
> >>> I've created a fairly fine surface of a sphere using the criteria:
> >>>
> >>>         30.0, // angular bound
> >>>         0.01, // radius bound
> >>>         0.001  // distance bound
> >>>
> >>> The surface is not smooth - it has an "orange peel" look.  Plotting the
> >>> radius of each vertex of the triangulation shows the magnitude of error
> >>> of the triangulation point positions from the expected radius of 1:
> >>>
> >>>     fineSphereSurfaceWithRadius.png
> >>>
> >>> Creating a coarse surface with criteria
> >>>
> >>>         30.0, // angular bound
> >>>         0.1, // radius bound
> >>>         0.1  // distance bound
> >>>
> >>> Results in:
> >>>
> >>>     coarseSphereSurfaceWithRadius.png
> >>>
> >>> where the magnitude of vertex error is similar, but in the fine case,
> >>> the same radial vertex position error in a smaller triangle creates a
> >>> bigger error in the normal direction, hence the uneven look.
> >>>
> >>> Is this the result that is expected? and is there anything that I can
> >>> do to improve the conformance of the triangulation points to the
> >>> expected sphere function?  Am I missing a tolerance adjustment
> >>> somewhere?
> >>>
> >>> It isn't a write precision/truncation problem, as I've set the
> >>> precision of the ofstream to 15.
> >>>
> >>> Thanks,
> >>>
> >>> Graham
> >>>
> >>>
> >>>
> >>>
> >>> -----------------------------------------------------------------------
> >>>-
> >>>
> >>>
> >>> -----------------------------------------------------------------------
> >>>-
>
> --
> Mariette Yvinec
> Geometrica project team
> INRIA  Sophia-Antipolis