I'd like to create a mesh from a point cloud generated by video tracking, ideally using python for some kind of prototype at least.

Initially I thought this is a fairly easy task, connecting the vertices, creating the faces, done ;) Then I've read that Screened Poisson Surface Reconstruction is currently the best approach. Although there is a nice github repository and code to play with, it's hard to understand the papers.

Q: Is there another clever implementation/approach you would recommend or is Screened Poisson Surface Reconstruction the way to go? Any suggestion how to implement that as simple as possible will be greatly appreciated.

  • $\begingroup$ Could you mention any other requirements? Do you need the fastest / most memory efficient approach, or the simplest one to implement to get on with prototyping? $\endgroup$ Dec 8, 2016 at 20:02
  • 2
    $\begingroup$ Maybe this question on stackoverflow helps: link $\endgroup$
    – wolle
    Dec 8, 2016 at 22:52
  • 1
    $\begingroup$ Have you tried using MeshLab? It's one of its use cases by design. $\endgroup$
    – IneQuation
    Dec 9, 2016 at 9:23
  • 1
    $\begingroup$ Performance doesn't matter. I think understanding the concepts and accuracy is more important. Sorry if the question is too broad, but I thought that any experiences from experts, how to dive in and which concept is worth it, would be helpful for me and potential future visitors @trichoplax $\endgroup$
    – p2or
    Dec 9, 2016 at 12:12

1 Answer 1


There is algorithm called as delaunay triangulation which does triangulation or points it's comparably simple to understand . But it is very slow. If you don't want to implement your own algorithm from scratch please have look at pcl http://pointclouds.org/documentation/tutorials/greedy_projection.php Also have look at CGAL http://www.cgal.org.

  • 2
    $\begingroup$ As a hint for the people wanting to learn this, the delaunay triangulation can be build with voronoi diagrams. A vornoi diagram takes an unordered set of points and calculates vornoi regions such that each region contains exactly one point and the border of a region is set such that the two points of the bordering fields have exactly the same distance to the border. Once the voronoi diagram is build, the delaunay triangulation is a simple matter of connecting each point with the points of the neighbouring voronoi regions. $\endgroup$
    – Tare
    Nov 28, 2017 at 14:03

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.