Stabilized Noise

Abstract

Perlin noise is generated by interpolation of a pre-defined random number table. However since the random number table is generated without considering the ideal properties of noise for computer graphics, the resulting noise function does not have these properties. We improve the properties of noise for Perlin s and similar algorithms, by stabilizing the random number table itself. We create the noise function using well-known statistical tools that measure the degree of stability of a random number table. These tools are used within an optimization procedure to create a random number table with a uniform random distribution, without periodicity, and having a band-limited property.