I'm a bit confused about what the definition of Perlin Noise actually is.

This article suggests that the Perlin Noise function refers to how different octaves of noise are added together, but it does not matter what method is used to generate the random number,

However the wikipedia page suggests that Perlin Noise refers to the actual noise function, while adding different octaves of noise creates fractal noise.

It seems to me that the first article is using the term Perlin Noise in the wrong way, I just wanted to check before I use it as a source for some work I'm doing. Thanks.

Hi,

Perlin Noise has nothing to do with the combination of different octaves.
Noise that is generated by adding up octaves is generally called FBM ( Fractal Brownian Motion ).
Althoug the article you refered is confusing about that, it is a good introduction to FBMs.

To create the FBM, you need some way to generate noise at different frequencies.
Perlin Noise is a method you can use to generate such octaves, it belongs to
the family of gradient noise methods.
Ken Perlins website introduces the algorithm.
Here you can find the reference implementation of Perlins latest algorithm.

However most tutorials use value noise to generate the octaves and falsely call it "perlin noise".
A tutorial that shows a simple value based noise is devmag.org.za/2009/04/25/perlin-noise/. ( the author calls it "perlin noise" but in fact it is simple value noise )
Value noise is faster to compute but has a lower quality.

Thanks for clearing that up for me guys, very helpful . Unfortunate to see so many resources get it wrong.