Approach of spatial sub sampling

[theohaschke]

Hey, i wonder what approach of spatial sub sampling cloudcompare uses ?
The documenation explains more the idea than the approach.
So does cloudcompare use a voxelgrid , compute the center point of all points in that vocel grid and take the closet point from original pointcloud of the center?
Or does cloudcompare determine an initial point and and from that point all further points must have a distance from the determined value?

I am trying to investigate several different approaches of spatial sub sampling, so I would be interested in a more detailed explanation.
Thank you in advance!

[daniel]

We flag all the points with a flag. Then, for all the points: if it still has this flag, we look at its neighbors and remove their flags. Otherwise we skip it and look at the next point.

Of course it depends a little bit at the order in which you look at the points, but overall and statistically it works pretty well (and it's super fast). In a way we just guarantee that no point is closer to any other than a given distance.