Generating blue noise point distributions with spectral analysis
The void-and-cluster algorithm (Ulichney, 1993) generates blue noise distributions with optimal spectral properties. Unlike white noise (purely random) which creates clusters, blue noise produces evenly spaced points without visible patterns.
The algorithm works by iteratively finding the "tightest cluster" (densest area) and "largest void" (emptiest area) using Gaussian-blurred density maps, then moving points from clusters to voids until the distribution converges.
Blue noise is widely used in: dithering for printing and displays, sampling in ray tracing and rendering, procedural content generation, and halftone screening.