Quote Originally Posted by mixerbach View Post
What makes O'Leary's generator different is that his program utilizes an erosion simulation created with real data.
My question would be "different than what?"

It's an interesting piece of work, though, especially the notes page that walks through the underlying parts. As the notes page points out, the basic generation follows the same pattern as http://www-cs-students.stanford.edu/...ap-generation/ and that page has a number of interesting links towards the bottom.

I'm more a regular grid sort of guy, myself. The artifacts are different (tending more towards 8-way than 6-way as with a well-relaxed noise grid), but some of the processing for a regular grid is nicely amenable to easy vectorization and parallelization, which is a little harder to do with a Voronoi-type of algorithm.

Florian Boesch had an interesting demo that runs a terrain erosion simulation on the GPU from your browser ( http://codeflow.org/entries/2011/nov...g-and-erosion/ ). The nice part of this code is that it uses a regular grid to get the same hexagonal connectivity that you get from a well-relaxed random grid while still keeping the regular grid structure that allows for GPU speed.