Terrain Generation: Infinite 2D space filled with Diffusion-limited aggregation clusters? Disclaimer: I don't have a deep understanding of fractals or any higher math, I'm just personally interested in it, so please excuse me if I'm using wrong terms or if I'm being inaccurate. Making things worse, I'm German ;)
Hi everyone,
I'm currently building a new terrain generator for Minecraft and I'm searching the web for the best and computationally cheapest methods to create nice looking terrain.
So far I'm using several octaves of Perlin noise overlaid, which gives me nice rough terrain, but it lacks the branchedness of real mountains.
Then I saw Diffusion-limited aggregation and it produces exactly the kinds of shapes I need, but so far I've only seen simulations in limited space.
My dream would be an algorithm that I can query with a specific position and seed (like Perlin noise) that infinitely produces pseudorandomly distributed DLA clusters.
Is this even possible?
If the above is impossible, do you think prerendering several seamless DLA tiles and overlaying them in different frequencies would produce random-looking results?
(I've found a CS homework assignment on the web for a program that produces seamlessly tileable DLA clusters.)
Thanks for your wisdom!
 A: Old question but I'll give it a shot: there are fractal algorithms (like the classic fern) where you can caculate at a given point, but the majority of these algorithms are too uniform to look correct.  You cannot calculate DLA at a given point because it is inherently based on a step-by-step process (at least, I could not think of a simple way to fake it).  You also cannot overlay DLA tiles if your aim is to produce realistic mountains -- The low frequency DLAs must fit exactly in their parent DLA patterns to look like correct mountain ranges with hydraulic erosion.  I am working on this problem right now, check my blog every so often I will probably have results up in the next few weeks.
A: 

You cannot calculate DLA at a given point because it is inherently based on a step-by-step process


Thats true, but DLA can be used as "seed". Its possible to generate terrains based on DLA. You can find an explanation of the algorithm on this blog post. These terrains are maybe not 100% realistic, but they look quite good and are simple to implement.

