Intelligent height calculation (must read)

uberzev, you nailed something that's been bugging me over and over, the height map looks ok but the results gets crappy.

Place this post in the Hall-OF-Fame!

I don't see how it is possible to autoprocess an arbitrary like you want - how should the algorithm know if you meant hemisphere and not cone (or visa versa)?

The solution is to intelligently construct the filter so various features and objects are placed correct in z dimension (and not stretched, like with your first height map ex.).

here is the profile for the second modified height map that produces the intended result:

Quote
Sphinx. wrote: how should the algorithm know if you meant hemisphere and not cone (or visa versa)?

I don't have the answer to that. But we did put a man on the moon so this should be easy.

Quote
Sphinx. wrote: I think the images explain well what is "causing" the bad shading, and why it is not a bug..

I realize it's not a bug. I just think an option for a more intuitive behavior would be awesome.

But it would be interesting with some sort of algorithm that scales features relative to their size. The main problem is locating and isolating separate features/objects. Think of a heigh map with mixed perlin noise and cell noise (blended).. how do you theoretically separate features here? It will appear as one "connected" surface object..

I once experimented with scaling the cell noise cells to get a more uniform shading (like what you're looking for). I did find a solution, but of course it only works for that given construction. Let me see if I can find it...