HN Explorer

10 comments

[ 5.7 ms ] story [ 40.5 ms ] thread
amelius
Isn't this a fancy name for just iteratively moving a vertex towards the center of mass of its connected vertices?
nkrisc
It’s a shorter name for that.
gsf_emergency_2
Truly fancy name would be something like, "Laplacian smoothing via gravitational accretion of vertetesimals"?
ramblingrain
It sounds like it doesn't have any guarantees regarding topology. Or maybe it's so simple or does?

Let's be positive my friends

itishappy
I don't think the algorithm as presented adds or removes new neighbors, preserving topology (except it won't prevent self-intersections using only neighbor positions). I imagine this is often used with some type of adaptive meshing algorithm for to make something like the clay editing model shown in the Substance demo.
nosferalatu123
(author here) That's correct. The algorithm preserves topology because no new vertices or indices are created. It can result in self intersecting meshes, but one way that this can be used is what Substance Modeler does, which is continuously remesh.
akomtu
Which is a fancy name for a low-pass filter?
ramblingrain
This reminds me of bouncing on a trampoline. I'm a human not a computer
peterdsharpe
Funny you should mention that! The surface of a trampoline can actually be modeled as a 2nd-order wave equation PDE, which uses a Laplacian spatial kernel - for the same reasons this is called Laplacian smoothing!