CGAL users discussion list

[Monique Teillaud <>]

Hi Claudia

except if the shape is particularly well sampled, there is usually no 
satisfying result when reconstructing with an alpha shape... it will 
either have holes or be "too large" or probably even both at the same 
time. As Mariette already sayd, this is not a good reconstruction method.

remarks about the size of your triangles:

- what you call "the" circumscribing sphere of a triangle in 3d is not 
uniquely defined, the alpha shape deals with one of the spheres passing 
through the 3 vertices of the triangle

- the max radius is sqrt(alpha) = 0.036... in your case
an edge length is at most the diameter = 0.072
so, your 0,055 is not surprising at all

best,
--
Monique Teillaud
http://www.inria.fr/sophia/members/Monique.Teillaud/
INRIA Sophia Antipolis - Méditerranée
Institut National de Recherche en Informatique et Automatique

Le 26/11/13 21:01, Claudia Werner a écrit :
> Thanks so far.
> I already "played" with different alpha-values but couldn't find a
> sattifying result. I either get holes in the bunny or the ears are
> connected by other triangles. Especially this (wrong) triangles are the
> triangles I am wondering about. As far as I understood, a facet of the
> Delaunay-Triangulation is part of the alpha-shape iff its circumsphere
> is empty and has a radius smaler than alpha. But the radius of the
> circumsphere of the triangles between the ears of the bunny have a
> radius mush greater than alpha, so why aren't this triangles deleted?
> E.g. when using alpha-opt = 0,0013... there are still triangles with
> edge-length around 0,055 otherwise mush smaler triangles are "deleted"
> Best,
> Claudia
>
>
>
>
> ----- Ursprüngliche Mail -----
> Von: "Daniel Duque" 
> <>
> An: 
>
> Gesendet: Dienstag, 26. November 2013 10:42:09 GMT +01:00
> Amsterdam/Berlin/Bern/Rom/Stockholm/Wien
> Betreff: Re: [cgal-discuss] Alpha_Shapes_3 Bunny
>
> Hi,
>
> I don't know much about the bunny, but if it's a nearly-regular mesh,
> with a well defined mean distance
> between surface nodes, d, you should go for alpha values around, but
> larger than, d^2.
>
> Best,
>
> Daniel
>
>
>
> On Tue, Nov 26, 2013 at 10:30 AM, Mariette Yvinec
> <
>  
> <mailto:>>
>  wrote:
>
>
>     Plain alpha_shapes is not a very good reconstruction method.
>     The "optimal" alpha-value is the smallest alpha value
>     that will give a single connected components including all data points.
>     It might not be the best one to see the Bunny feature appear.
>
>     You might try a few others alpha values,
>     just to be convinced things are working as they should.
>
>     Le 25/11/13 20:28, Claudia Werner a écrit :
>
>         Dear all,
>         I am working on Alpha_Shapes for triangulating a point cloud. As a first 
> example I tried the example "Alpha_shapes_3/ex_alpha_shapes_3.cpp"
>         
> (http://doc.cgal.org/latest/Alpha_shapes_3/Alpha_shapes_3_2ex_alpha_shapes_3_8cpp-example.html)
>  and wrote the result to an off-File as descirbed 
> here:http://stackoverflow.com/questions/15905833/saving-cgal-alpha-shape-surface-mesh.
>         Displaying the result with MeshLab (using the off-file), I was 
> wondering about the result (see attached picture).
>         Has anybody an idea why I don't get a better result? As in the 
> example I used the optimal alpha.
>         I also attached the cpp-file.
>
>         Best Regards and thanks for your help,
>         Claudia
>
>
>     --
>     Mariette Yvinec
>     Geometrica project team
>     INRIA  Sophia-Antipolis


--
Monique Teillaud
http://www.inria.fr/sophia/members/Monique.Teillaud/
INRIA Sophia Antipolis - Méditerranée
Institut National de Recherche en Informatique et Automatique