CGAL users discussion list

["Sebastien Loriot (GeometryFactory)" <>]

Brad wrote:
> Hi,
>
> I have a set of 3D points & normals, but they are distributed throughout a
> volume, not just on a surface. Is the class Poisson_reconstruction_function
> likely to be able to handle this?
I would say no. Points inside the volume bounded by the surface you
expect will influence the output surface (if it can compute one).
> What I'd like to do is: add all the points, call compute_implicit_function,
> and
> use operator() to get the integrated values at the point locations. I don't
> (yet)
> require a surface mesh.
You can give it a try and see what it gives.
If you are able to have a method to roughly filter point inside and on
the surface (even almost on the surface), it might give better results.

S.

> The documentation says that the implicit function is zero on the surface (at
> the points, if the points are sampled on a surface) and negative inside. Is
> there something in the algorithm that enforces zero value at the points?
> Because
> only the exterior ones in my point cloud should have value zero, the
> interior
> ones will be the nonzero result of the poisson integration.