Hi @nktlst,
If you decide to use Delaunay triangulation I would suggest to either:
- rotate slightly the triangles (not too much to avoid overlap)
- sample 3 equidistant points along the edges of the current triangles in order to create new offset triangles
This way you can break the symmetry inherent to a Delaunay triangulation (adjacent triangles have parallel edges) and have something that resembles your example picture.
Another thing to consider is the placement of your points. If you want the triangles to be more or less equidistant it may be useful to resort to Poisson Disk Sampling before triangulation.
Alternatively, you could:
- compute a Voronoi diagram out of your point cloud
- randomly scale down the Voronoi cells
- sample 3 equidistant points along the edges of the cells to create a triangle
Then no need for rotation or extra-sampling since the dis-symmetry will appear naturally.
Of course that’s just one way of doing things. You could also use brute force methods with collision detection or resort to simple physics engines with rigid bodies and repulsion forces.