But I don't think, a video codec would work. Or maybe just with very few points/triangles. Depending on the point-selection (equally distributed is a lot faster!) it takes some time between 50s to 5 minutes for a triangulation of about 4000 Points.
For just a few hundred it could work, but I don't think it's good enough for real-time.
(Displaying is no concern. You can render thousands of triangles in real-time)
I think it's part time the language choice and the fact, that my algorithm probably runs in Omega(n logn) but might not reach O(n logn)...
Maybe you got an idea: My bottleneck is to find the triangle in which a new point lies. Right now I go through the triangle-mesh in the direction of the point, hoping to find the right triangle fast. Most of the time, it does ...
I've wondered what a video codec for this would look like, compute the triangulation for I frames and in between just tweak the location of the points, when triangulation is totally broken, time for a new I frame. It's somewhat slow to compute, but displays well on hardware optimized for displaying triangles...