Interesting Failures

Today's post is going to show the results of a few different approaches I've tried to use for generating terrain.  None of these quite hit the mark, but I found them visually interesting.  The first two examples are for entire planets.

In the first one, I started out by creating a supercontinent using a probablistic fill algorithm.  Next I split it into continental chunks and shifted the chunks, then added mountains along the medial axis of the resulting chunks.  Interesting, and you can see how they once fit together (like South America and Africa once did), but not convincing.

The second one was a variation on the theme, but more colorful.  Here, I generated the supercontinent using an algorithm that combined a decay equation with Perlin noise.  The supercontinent was split into chunks ago, which where then randomly placed in cluster zones.  A dew smaller chunks were spread randomly, and island arcs were generated.  The puzzle-piece effect of the first one is missing, but the coast lines here are more interesting.  Still not a convincing map to me, though.

The third image results from a bug when porting Amit Patel's Polygonal Map Generator from Flash/ActionScript to Windows Forms/C#.  (I should clean that up and get it up on GitHub someday.)  That program is intended to generate not the terrain for a whole planet, but only for an island.  Alas, a bug while testing the port resulted in the image below, which reminds me of some type of four-legged animal leaning down to eat or drink something.






















That's it for today.  I'll probably be posting more of Pandora's Gate tomorrow.