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 main blue noise sampler (RGBN) implemented here have linear complexity in the number of points and the dimension, and run in under 15 minutes for 1 million 2D points.
📦 Installation
pip install blue_sampler
🚀 Quick start
import blue_sampler as blue
# 10 000 points in 2D
x = blue.sample_points(N=10_000, D=2)
blue.plot(x, auto_zoom=True)
# structure factor visualization
blue.plot_structure_factor(x)
# higher dimensions
x = blue.sample_points(N=2_000, D=5)
# image stippling
x = blue.im2points(image="zebra.jpg")
🖼️ Example
📊 Supported dimensions
| Dimension | Notes |
|---|---|
| 2D | Fast, recommended |
| 3D | ~2× slower |
| 4–5D | Works, more iterations needed |
| ≥6D | Experimental (small N recommended) |
📚 Links
- 🌐 Project website: https://for-a-few-dpps-more.github.io/rgbn/
- 📦 PyPI: https://pypi.org/project/blue-sampler/
- 🐙 GitHub: https://github.com/For-a-few-DPPs-more/rgbn
- 📖 Documentation: https://blue-sampler.readthedocs.io
📄 License
MIT