CGAL users discussion list

[Hamid G <>]

Thanks Sascha, this was a great help.

Do you know if the seeded points are going to be preserved in the mesh
or they are just temporary guides for the meshing algorithm?



On Fri, Sep 28, 2012 at 12:15 PM, Sascha Zelzer
<>
 wrote:
> Sure,
>
> I have attached a file with my code. You can ignore the types in the "mitk"
> namespace, or replace them with CGAL types where appropriate.
>
> In my case, I have a labelled image as the input for the CGAL mesher for
> which I can provide a mapping of labels to a sampling distance for the seed
> points (the SeedPointDistanceToLabelMap type in the attached code). So for
> example I know that my image contains three labels, with consecutive values
> from 1 to 3. For value 1 I need a sampling distance of 15mm (for example),
> for value 2 5mm, etc. I then simply iterate over all different sampling
> distances and for each distance, I iterate over the whole image (with
> stepping according to the current distance) and create a seed point if the
> current image value corresponds to a label value for which that sampling
> distance is required.
>
> The created seed points are then inserted into the triangluation object of
> my domain (instead of relying on the default constructed seed points in
> make_mesh_3). I'm not sure if that approach has any disadvantages (apart
> from brute force iterating over the image) but it seems to work okay.
>
> - Sascha
>
>
> On 09/28/2012 05:47 PM, Hamid G wrote:
>>
>> Sascha,
>>
>> Would you mind elaborating on what exactly you did? I have same
>> problem and using varying sizing field didn't help me either.
>>
>> thanks in advance,
>>
>> On Mon, Sep 17, 2012 at 3:37 PM, Sascha Zelzer
>> <>
>>  wrote:
>>>
>>> Hi,
>>>
>>> I was finally able to construct a suitable set of seed points and run the
>>> mesh generation step successfully.
>>>
>>> Many thanks for your hints!
>>>
>>> Before my first posting to this list, I have tried using a
>>> CGAL::SizingField
>>> and setting different values for the tet radii for the different
>>> sub-domains. That did obviously not work out and I now know it is because
>>> of
>>> an inappropriate default seed point set.
>>>
>>> Thanks,
>>>
>>> Sascha
>>>
>>>
>>> On 09/10/2012 09:57 AM, Mariette Yvinec wrote:
>>>
>>> Currently,
>>> the best way to ensure that small component are mot missed
>>> is to set initial points in them or on them boundary.
>>> Then you have to use refine_mesh_3 instead of
>>> make_mesh_3 for your initaial seeding points to be taken into account.
>>>
>>> Another way, if you know the area where the small components are supposed
>>> to
>>> be,
>>> is to define a  variable sizing field with a value locally similar to the
>>> size of the components
>>> to be meshed.
>>>
>>>
>>>
>>> Le 30/08/12 16:18, Sascha Zelzer a écrit :
>>>
>>> Hi,
>>>
>>> I have posted about my problem already here
>>>
>>> http://cgal-discuss.949826.n4.nabble.com/3D-volume-mesher-ignoring-small-subdomains-td4655756.html
>>> but thought I reformulate it:
>>>
>>> I have a labeled image as input to make_mesh_3 which contains a small
>>> sub-domain (label 2) inside a larger domain (label 1). Small meaning the
>>> its
>>> characteristic size is about two orders of magnitude smaller then the
>>> embedding domain.
>>>
>>> So my question is, what are the criteria for refining cells in the large
>>> domain to ensure that they do not "shadow" a complete small sub-domain? I
>>> would like to be able to set a large maximum cell size for the large
>>> domain
>>> and a suitable cell size for small sub-domain. When setting the cell and
>>> facet criteria in such a way, my small sub-domain is ignored however.
>>>
>>> Any tips would be highly appreciated.
>>>
>>> Thanks,
>>>
>>> Sascha
>>>
>>>
>>> --
>>> Mariette Yvinec
>>> Geometrica project team
>>> INRIA  Sophia-Antipolis
>>>
>>>
>>>
>