Processed deep-sky astrophotography image of Arp 319, Stephan's Quintet

Noise Off. Photons On.

Scientific image-processing tools

Open-source tools by Benoit Blanco for PixInsight, Hubble, JWST and deep-sky astrophotography. Preserve astronomical signal, remove instrument artifacts and inspect every correction.

Processing guide

Hubble and JWST image processing in PixInsight

Start from the physical defect, then choose the appropriate correction for cosmic rays, detector lines, Hubble drizzle stripes or random noise. The guide separates validated Hubble workflows from the more cautious use of tools on JWST products.

Open the guide
Compatibility update

Ready for PixInsight 1.9.4 and Linux

BB-Astro scripts now use the V8 JavaScript runtime required by PixInsight 1.9.4. The update repository installs the current V8 release on Apple Silicon, Intel Mac and Linux, while keeping a compatible historical release available for PixInsight 1.8.0 through 1.9.3.

PixInsight 1.9.4+ Apple Silicon Linux x64 Legacy releases retained
Install the scripts
Oriented Hubble Destriping

StripeField

Model and subtract weak detector row-bias fields after drizzle has reprojected them to arbitrary angles. StripeField protects sources, never rotates the science image, and opens the exact signed field for quality control.

Explore StripeField
Arp 255 Hubble mosaic after StripeField correction Arp 255 Hubble mosaic before StripeField correction
Input Corrected
Cosmic Ray Rejection

LACosmic

The gold standard for cosmic ray rejection. Implements van Dokkum’s algorithm with astroscrappy. Laplacian detection, PSF-aware star protection, and iterative cleanup for HST/JWST-class data.

Explore LACosmic
Cleaned Raw
Raw Cleaned
Deep Learning Removal

DeepCR

Next-generation cosmic ray removal using deep learning. Trained on 15,000+ HST images for superior detection on 32-bit float images and space telescope data.

Explore DeepCR
Arp 204 Hubble image after DeepCR cosmic-ray removal Arp 204 Hubble image before DeepCR cosmic-ray removal
Raw Cleaned
Denoising with no training data

Deep Image Prior Linear

A randomly-initialized neural network is fitted to your single linear image and stopped just before it starts learning the noise. No training set, no pre-trained model: nothing can be hallucinated from data it has never seen. Full frame, photometry preserved, validated on Hubble fields.

Explore DIPL
Arp 176 satellite galaxy after DIPL denoising, 1:1 crop Arp 176 satellite galaxy before denoising, 1:1 crop
Input DIPL
The telescope that denoises itself

AstroSURE

A Noise2Noise U-Net trained on pairs of dithered Hubble exposures of the same field: the two frames share the signal but not the noise, so the signal is the only thing the network can learn. No clean images anywhere in the training, seconds per frame, photometry validated on Gaia DR3.

Explore AstroSURE
Arp 70 galaxy after AstroSURE, 1:1 crop Arp 70 galaxy before denoising, 1:1 crop
Classic AstroSURE
Complete collection

Free PixInsight scripts for astronomical image processing

BB-Astro tools cover cosmic-ray removal, hot and cold pixels, defective detector lines, Hubble destriping and linear-image denoising. Each script has a documented method, practical settings, comparison images and explicit limits.

LACosmic and DeepCR

Remove cosmic-ray hits from Hubble, space-telescope and long-exposure astrophotography data.

Getting Started

Where to find the scripts

Once installed, you can access the modules directly from the PixInsight menu bar under Script > BB-Astro.

PixInsight Script Menu Location