2
\$\begingroup\$

I am experimenting with procedural content generation, and starting with generating terrain from fractional brownian noise. The terrain produced looks fairly realistic, but has many artifacts and splotches of random terrain everywhere. Here are some examples: enter image description here enter image description here enter image description here

How can I clean up the terrain so that it has smooth transitions from terrain type to terrain type without the scattered spots?

|improve this question
\$\endgroup\$
  • 1
    \$\begingroup\$ what are the undesired spots and what have you tried to "smooth" it out? \$\endgroup\$ – wes Aug 23 '13 at 14:39
  • \$\begingroup\$ @wes - The undesired spots are the splotches of different colors that appear out of nowhere. For example, in the last image in the bottom left corner there is that splotch of dark green, and in the first image in the mid/upper/right above a sandy area is a small isolated patch of dark green. I have a smoothing function that smoothes out the noise to generate the terrain as it is now, but applying it further for some reason degenerates the terrain. \$\endgroup\$ – pasawaya Aug 23 '13 at 14:41
  • \$\begingroup\$ is it the size that makes those splotches undesirable? if so, you can process your data again and if the patch doesn't cover a certain area, replace it with something more desirable \$\endgroup\$ – wes Aug 23 '13 at 14:50
  • \$\begingroup\$ honestly though, this is how perlin is supposed to work, it wouldn't "appear" to be so natural/random if it were more uniform. you can increase or decrease the bounds checking that determines what color splotches to place to try to smooth it out, or you can "errode" it after generation \$\endgroup\$ – wes Aug 23 '13 at 14:51
  • 2
    \$\begingroup\$ You should tell us something about your implementation of fractional brownian noise. By definition it is the sum of several octaves of noise. Octave means a multitude of a predefined frequency. The contributions of the highest frequency (aka smallest wavelength) create the smaller feature. So maybe limit the number of octaves. \$\endgroup\$ – SanityIsOverrated Aug 23 '13 at 17:04
2
\$\begingroup\$

is it the size that makes those splotches undesirable?

if so, you can process your data again and if the patch doesn't cover a certain area, replace it with something more desirable

|improve this answer
\$\endgroup\$
  • \$\begingroup\$ Adding some bleeding could destroy the smaller area. \$\endgroup\$ – DogDog Aug 23 '13 at 15:45

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.