[rohitarondekar]

This is just a simplistic description of Perlin Noise. However while implementing the algorithm this is not the way to go. Because it's very very slow -- the author has simply taken non-coherent noise and smoothened it and this is computationally very expensive.

And also Ken Perlin has made an improved version to his original with some speedups and a reference implementation in Java can be found here: http://mrl.nyu.edu/~perlin/noise/

[bootload]

"... I first started to think seriously about procedural textures when I was working on TRON at MAGI in Elmsford, NY, in 1981. ..." ~ http://www.noisemachine.com/talk1/4.html

there is another site Perlin has explaining how, "Making Noise" ~ http://www.noisemachine.com/talk1/

[imd]

According to https://secure.wikimedia.org/wikipedia/en/wiki/Value_noise (which cites this), this page actually describes value noise, not Perlin. I emailed the author, who told me he no longer has access to the site, so unfortunately, people will be fooled for years to come.

[J3L2404]

Where are the permutation numbers derived from?

[rohitarondekar]

They are generated randomly. Ken Perlin has described it in his book Texturing, Modeling A Procedural Approach. If I remember correctly the numbers have to be such that every direction should have an equal chance. Even distribution I think.

[J3L2404]

I was thinking it would be more like the fast inverse square from Quake. If it's random, using the same code could be a problem, no?