0

I know how to generate a Voronoï / cell noise such as this one using Delaunay Triangles :

Cell noise

But how do I apply noise to the lines to make them more natural ? I cannot have sharp edges for procedural generation as it would look very out of place and unpleasant.

I am looking for a result that would somehow look like this :

enter image description here

( the picture is from a more advanced project )

Note : I cannot generate the entire map at once ( it is too big ) so the Voronoï diagram is used as metadata but I need a way to know in what cell are the coordinates (x, y) after deformation in order to make it work.

Paul Floyd
  • 5,530
  • 5
  • 29
  • 43
Matthieu Raynaud de Fitte
  • 2,001
  • 3
  • 25
  • 49
  • I think you have to divide the lines with smaller lengths and add normal distribution noise to all the points. – artgb Oct 23 '17 at 01:49
  • how would making the lines smaller help me ? – Matthieu Raynaud de Fitte Oct 23 '17 at 01:50
  • 1
    Are you using library or building your own? anyway you can get all the points of voronoi, Then add points to long length lines(divide lines with smaller than 3). – artgb Oct 23 '17 at 01:53
  • doing everything on my own ( including Perlin noise ). Are you saying that I should devide in smaller segments and then start moving the points that define them in ordre to get a more "broken" cell border ? – Matthieu Raynaud de Fitte Oct 23 '17 at 14:03
  • Yes, if that's right – artgb Oct 23 '17 at 14:06
  • would it no be better to apply noise to the line in order to get something very detailed ? On big maps it would mean a very big quantity of points. I will use your method if I cannot use noise however. Can you write the tip as an answer so I can leave an upvote ? – Matthieu Raynaud de Fitte Oct 23 '17 at 14:42

1 Answers1

1

I would randomize 3 - 5 points on each line to generate sub segments, based on a seed computed thanks to the coords of the two original segment points.

This kind of random seed allows to get the same results each time. You could thus cache the results or decide to compute the same ones again.

Maybe more zoom means more random sub-segments based on the same method.

Christophe Roussy
  • 16,299
  • 4
  • 85
  • 85