blue-sampler 0.1.10

Stealthy point-pattern sampling on the unit torus

blue-sampler

Generate large stealthy point patterns on the unit torus [0, 1)^D.

Stealthy point patterns have vanishing density fluctuations at low ("blue") frequencies, making them useful for Monte Carlo integration, image stippling, and any application that needs well-spread, low-discrepancy points.

The blue noise sampler implemented here have linear complexity in the number of points and the dimension (per gradient step), and run in under 10 minutes for 1 million points 2D points.

Installation

pip install blue_sampler

Quick start

import blue_sampler as blue

# 10 000 points in 2-D
x = blue.sample_points(N=10_000)
blue.plot(x)
blue.plot_structure_factor(x)

# arbitrary dimension D
x = blue.sample_points(N=2_000, D=5)

# image stippling
x = blue.im2points(image="zebra.jpg")

Supported dimensions

D	Notes
2	Fast, recommended for exploration
3	~2x slower than 2-D
4–5	Requires more iterations
≥ 6	Experimental, for small sample sizes

License

MIT