CGAL users discussion list

[Michael Bieri <>]

Unfortunately, the proposed solution with the nonlinear growth does not really work. It seems to be better in some parts, but the extremely large number of small elements does not go away.

(In the meanwhile I just wrote a mesher from scratch which seems to serve my purpose. During that work it seemed that it as well has trouble at region boundaries with a rather acute angle (="spatially slowly coming closer"), like for example when two ellipses are placed next to each other. As soon as the enclosed "outer" region becomes smaller than the wished element size, it has no other choice to further and further shrink the elements. So I smoothened the boundaries to blur out geometric features smaller than the element size. As soon as time permits, I'll try wether that solves the problem with CGAL.)

On Thu, Oct 5, 2017 at 1:09 PM, Michael Bieri <> wrote:

Ok, I will try that.

But what I don't get: Why does it need so tiny elements to approximate the 1D features in some areas of the mesh, while it is perfectly fine with pretty large elements around 1D features in other areas?

On Thu, Oct 5, 2017 at 1:03 PM, Laurent Rineau (CGAL/GeometryFactory) <> wrote:

Le Thursday, October 5, 2017 11:46:16 AM CEST Michael Bieri a écrit :

> I just uploaded the files to the Dropbox folder:

> https://www.dropbox.com/sh/lz8xmwhr3fg23qu/AAAZEpWnKzJyrjhNB006_-fVa?dl=0

>

> I further simplified it a bit, there's now just one region wit the

> ellipses. But the effect remains the same.

>

> There are the following files:

> - 01 - 04.png: The old images

> - dist.png, region.png: 2D view of the distance field and region supplied

> to the meser

> - img_elem_size.inr, img_regions.inr: 3D images for CGAL mesher

> - mesh: The mesh

> - Screenshot***: A screenshot of the mesh.

> - Furthermore, the source code. It's a bit of a mess because it was written

> and edited very quickly...

 

There is nothing unexpected in the result you got.

 

I propose a tiny patch, that uses a non-documented feature... Can `make_mesh_3` that way:

 

CGAL::parameters::internal::Mesh_3_options options;

options.nonlinear_growth_of_balls = true;

C3t3 c3t3 = CGAL::make_mesh_3<C3t3>(domain, criteria, options);

 

Then the 1D-meshing of curves will be dense were it is supposed to be.

 

--

Laurent Rineau, PhD

R&D Engineer at GeometryFactory http://www.geometryfactory.com/

Release Manager of the CGAL Project http://www.cgal.org/