Fractional Brownian motion (fBm)
Summing several octaves of a base noise — each at higher frequency and lower amplitude — into self-similar fractal detail for terrain, clouds and textures.
A single octave of Perlin or simplex noise gives smooth variation at one scale; fBm layers several. Each successive octave multiplies the frequency (the lacunarity, usually 2) and scales down the amplitude (the gain or persistence, usually about 0.5), so broad hills get fine wrinkles without a new function. Four to eight octaves is typical.
The result has the self-similar look of real terrain, cloud cover and marble, and it is the backbone of procedural heightmaps. The name comes from Mandelbrot and Van Ness's 1968 stochastic process, but the graphics version is just the weighted octave sum — often seasoned with domain warping or a ridged variant that flips valleys into sharp crests.