search

cnr-isti-vclab / meshlab

The open source mesh processing system
http://www.meshlab.net
GNU General Public License v3.0
4.69k stars 821 forks source link

Ball pivoting mesh reconstruction doesn't consider point/vertex normals #420

Open clausmichele opened 5 years ago

clausmichele commented 5 years ago

When using Ball Pivoting to reconstruct a mesh from a Point Cloud, the algorithm recomputes the normals, leading to a wrong direction in some parts of the mesh (see image). Recomputing face normals or inverting them does not solve the problem. Since I have to apply a texture to the generated mesh, this generates artifacts (see image). The point cloud has correctly oriented normals, computed setting the point of view.

Point cloud with correctly oriented normals: immagine

Generated mesh with Ball Pivoting with applied texture: immagine

Hugal31 commented 5 years ago

Reproduced on v2016.12

mcallieri commented 5 years ago

yes, ball pivoting only uses coordinates. The original algorithm was developed to triangulate sparse, non-uniform clouds of points without normals, and that is the version we have in MeshLab, at the moment. It could be nice to take in account also normals (maybe controlled with a parameter), but I have no idea when it will be possible to add this.

Hugal31 commented 5 years ago

This is strange, because the description of the algorithm states "Given a point cloud with normals it reconstructs a surface using the Ball Pivoting Algorithm."

stale[bot] commented 4 years ago

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.