JWST-StarClustering takes a single James Webb deep-field frame — Stephan’s Quintet — and pulls it apart into three populations: stars, dust, and galaxies. The guiding idea is that the structures worth isolating are separated not by color, but by size, so the whole pipeline is built on spatial-frequency filtering rather than thresholding.
The idea
In a dark-sky JWST frame, stars are compact point sources, dust is mid-scale clumpy structure, and galaxies are large luminous bodies. Those are three different scales, which means scale-space filtering can separate them cleanly where a single brightness threshold would just smear them together.
The pipeline
The image is lightly denoised with a 3×3 Gaussian blur (σ=0.5), then converted RGB → HSV so the work happens on the Value (brightness) channel, which stays stable across dark sky. Each population is then isolated with a different band of spatial frequencies:
- Stars — high-pass via unsharp masking:
Value − blur(σ=8), keeping only the sharpest point sources. - Dust — band-pass Difference-of-Gaussians:
blur(σ=18) − blur(σ=1), for mid-scale clumps. - Galaxies — wider band-pass:
blur(σ=18) − blur(σ=100), for the large, slow structure.
The result
The single input frame separates into three clean layers: stars carrying JWST’s characteristic six-point diffraction spikes, the mid-scale dust of star-forming regions, and the five galaxies of Stephan’s Quintet resolved as distinct blobs. Each stage is written out so the decomposition is reproducible end to end.
What I learned
This is exactly the front-end stage of professional astronomy pipelines — source separation and background estimation, the same job tools like SExtractor do at NASA/JPL and Caltech/IPAC. The lesson that stuck is that principled, reproducible parameter choices matter more than clever ones: pick your scales deliberately and the structure falls out on its own.