Adding interest to low-res DEM's

General Discussion
1

Hi Guys,
I’m trying to ‘realistically’ add detail to low res (90m per pixel) DEM based heightmaps, but I’m having trouble keeping the underlying lay of the land relatively untouched. Any pre-built macros out there that I can try? Or have any of you guys successfully attempted something similar?

thanks

.sub

2

There’s a few different possibilities to do this… some other folks can probably chime in to help also.

The options basically are :

  1. Add fractal noise to the terrain, making sure that the scale and amplitude of the noise is appropriate to not blow out your existing data.

  2. Use the displacement device to create detail/noise in your terrain without actually adding “new” data to it. Drive the displacement with a basic perlin noise, set to a small scale, and use only a small amount of displacement.

  3. Use the erosion device to add some fine-scale structure.

I often use all 3, but #2 and #3 in particular I find very useful

3

It’s funny you should ask that because this is something that I’m trying to do at the moment. Nikita may have some answers here too. The way I see it you’re trying to add mid scale and small scale noise. Anything larger would destroy the underlying dem. I tried chaining two erosion devices together: see the testing forum thread : ‘slowdowns’ for a link.
The idea was to first introduce the new levels of structure with some controlled erosion. Then with the second erosion pick it up from there and really create all of the recognisable filigree details. I managed to preserve out the river channels- sediment carry param is important.
It’s probably better to use a noise device to achieve the structural noise. I’m thinking Spectral Combiner…I’m not sure why. Lots to try myself including the perlin-displacement idea.

monks

4

I’d use a simple network like this one: http://nikita.tac-design.net/sonstiges/sample2.tmd
The Clamp Devices make sure that the sum isn’t greater than 1.
The combiner adds the perlin to the low-res heightfield.

5

Nikita, the sample2 tmd looks like it’s doing roughly what I was seeking to do. I see it’s using the same approach as you’ve used elsewhere.

monks

6

Because often it’s the best idea to start out with something simple. :slight_smile:
And as a beginner, simple networks are easier to understand.