Protect sources
An iterative median and MAD mask excludes galaxies, stars, spikes and other bright signal.
What it corrects
Striping after drizzle reprojection
Hubble ACS/WFC row-bias variations are horizontal in detector coordinates. Drizzle can reproject those detector rows to arbitrary angles in a final mosaic. Their amplitude may sit below the pixel noise, but their coherence across thousands of pixels makes them visible after a strong display stretch.
StripeField estimates a signed additive value at every pixel. It is not a defective-row selector and it never rotates the input image.
Scientific model
A field you can inspect
The correction is exactly corrected = input - stripe field. The full-precision
signed field opens in a separate PixInsight window. Apply an STF and verify that it contains
line structure only before keeping the correction.
If the model contains a galaxy, halo, tidal feature, star or diffraction pattern, reject the result and use more conservative settings.
Arp 255 HST example
Before and after
Drag anywhere across the image to compare the same strongly stretched display before and after correction. The processing itself was performed on the linear mosaic.
- Five oriented stripe families removed automatically
- Sources and extended emission protected by the mask
- No rotation or resampling of the science image
Input
Corrected
Quality control
The signed removed field
This is the model removed from the Arp 255 mosaic, displayed with its own strong stretch. It contains the coherent line structure and no visible imprint of the galaxy or stars.
The model window is enabled by default because this visual check is part of the method, not an optional presentation feature.
How StripeField builds the correction
Find directions
A noise-corrected scan tests angles from -90 to +90 degrees and refines the best candidate to 0.125 degrees.
Estimate profiles
Robust transverse medians isolate row-scale structure while broad astronomical background is removed from the estimate.
Shrink and subtract
Wiener shrinkage suppresses unmeasured components, then the accepted field is evaluated on the original pixel grid.
A concentration guard catches isolated satellite trails or diffraction spikes that could mimic a stripe direction. Those pixels are masked instead of being extended into a full-frame subtraction.
Validated defaults
| Parameter | Default |
|---|---|
| Maximum passes | 12 |
| Stopping significance | 5 sigma |
| Background smoothing | 100 px |
| Source mask threshold | 3 MAD |
| Transverse detrending | 25 px |
| Along-stripe component | Disabled |
Validated results
| Hubble mosaic | Families | Runtime |
|---|---|---|
| Arp 255 | 5 | 282 s |
| Arp 314 | 8 | 411 s |
| Arp 141 | 11 | 522 s |
The synthetic regression preserves
input = corrected + model to 1.86e-9 maximum error and reduces the
row-profile high-frequency RMS by 73 percent.
Use it on linear Hubble data
- Open the linear monochrome mosaic.
- Launch Script > BB-Astro > StripeField.
- Start with the validated defaults.
- Keep Open the signed stripe model enabled.
- Run the correction and inspect the model with an STF.
Do not run StripeField on a strongly stretched image. Stretching changes the noise statistics used by the mask, angle score and Wiener weights.
Current scope
- Linear monochrome Hubble mosaics
- Weak row-bias fields at arbitrary projected angles
- PixInsight 1.9.4 or later
- macOS or Linux with Python 3.10 or later
JWST is not validated. Its final mosaics can combine zone-dependent stripe families, detector footprints and exposure seams. The current global Hubble field should not be treated as a JWST correction.