Request: accelerated voronoi sampling using new SurfaceVoronoi algorithm

[zoyt]

Hello!

I use of the Voronoi Sampling filter, through PyMeshLab quite a bit for various projects, particularly the Quality Weighted functionality. It is fantastic, and yields incredibly helpful results. However, I frequently find myself working with a model where applying the filter will take 2-3 hours on my (fairly powerful) PC. Admittedly, the meshes are quite large (~250 000 verts), but I started looking into wheather there were any recent developments in 3d surface voronoi algorithms that might be faster, and found this paper and associated repo.

This is not my field of expertise, but my understanding of the paper is that their algorithm is substantially faster than others currently being used (particularly for large meshes), and allows for different distance solvers, akin to the Quality Weighted functionality. Frankly, I don't know what it would take for this to be integrated into meshlab/pymeshlab, but myself (and my electricity bill!) would be most appreciative!

Thank you.

[zoyt]

Returning to this after doing some testing - the Voronoi Sampling filter appears to be very sensitive to meshes with high aspect ratio (right term?) faces. In some areas where the triangles were almost needle-like without me noticing, the algoritm completely broke down and created cells that were 10x the area as others. Not certain what causes this, but the solution I found was to run the Remeshing: Isotropic Explicit Remeshing filter first to make the faces more equalateral. Now, the script that used to take several hours for mediocre results can be run in 10-15 minutes with much better results. While the algorithm in that paper is intersting, the situation is far less dire than I originally thought.

It may be useful for new users if there was a pop-up when applying this filter on a non-refined mesh, or even if a simple warning of this behavior was part of the filter documentation.