Using Scalloped Sectors to Generate Poisson-Disk Sampling Patterns

Sampling distributions with blue noise characteristics are widely used in computer graphics. Although Poisson-disk distributions are known to have excellent blue noise characteristics, they are generally regarded as too computationally expensive to generate in real time. We present a new data structure that alllows sampling by dart-throwing in O(N logN) time. We also show how a novel and efficient variation on this algorithm can be used to generate Poissondisk distributions in O(N) time and space.