Gradient Hunter

Gradient Hunter before/after gradient removal
Gradient Hunter - background model and corrected image for a light-polluted deep-sky field.

Gradient Hunter is AIT’s dedicated background modeling and gradient removal engine. It is designed to handle light pollution, moon glow, vignetting and other large-scale gradients on big sensors, while preserving natural colors and not chewing into your nebula or galaxy.

Quick Start (5-10 min)

  1. Load a stacked image (calibrated lights with darks/flats/bias already applied). Prefer a mostly linear or gently stretched version.
  2. Open Gradient Hunter from the Tools menu or from your workflow after stacking.
  3. Choose a tiling/grid size appropriate for your sensor and gradient complexity.
  4. Optionally define or enable a mask to protect bright nebula or galaxy regions from influencing the model.
  5. Run the background model and inspect the before/after preview and the model itself.
  6. If the result looks natural (smooth background, preserved colors), apply and save the corrected image.

Controls

Control Default / Range What it does When to use
Tile size / grid Medium Controls how the image is divided into patches for background estimation. Use smaller tiles for complex, local gradients; larger tiles for smooth gradients and noisy data.
Background statistic Median / robust Determines how the background level is estimated in each tile. Robust statistics help reject stars/nebula; mean may be useful in clean regions.
Sigma clipping On Rejects bright outliers (stars, hot pixels) while computing the model. Keep enabled in most cases; increase iterations or thresholds for very crowded fields.
Model smoothing / order Medium Controls how smooth the background model is across tiles. Higher smoothing for simple gradients; lower smoothing if gradients change quickly.
Preserve color balance On Applies the model in a way that minimizes global color shifts. Keep on for broadband data; only disable for very specific workflows.
Show model / Show difference Off Toggles between the background model, the corrected image and the difference image. Use to understand what is being removed and to spot over-subtraction.
Export background model - Saves the modeled background as a separate image. Useful for documentation and advanced workflows where you combine models manually.

Example workflows

Light-polluted broadband data

  1. Stack your lights with Live Stacking PRO or your preferred stacker.
  2. Open the linear or gently stretched stack in Gradient Hunter.
  3. Use a medium tile size and robust statistics with sigma clipping.
  4. Inspect the model; if it captures the sky glow structure, apply and proceed to stretching and color work.

Wide-field vignetting cleanup

  1. After flat calibration, some residual vignetting may remain, especially on fast lenses.
  2. Use a larger tile size so the model follows the broad vignetting pattern.
  3. Check that the model is smooth and does not show imprints of nebulae or the Milky Way band.

Tips & notes

  • Do gradient removal before heavy stretching and before aggressive color work.
  • When in doubt, err on the side of milder correction-a tiny residual gradient is easier to live with than a hollowed-out nebula.
  • Protect bright, extended objects with a mask if the model starts to carve into them.
  • If the background looks unnaturally flat or plasticky, reduce model strength or complexity.

Troubleshooting

  • Background becomes too dark or clipped → reduce strength, check that black point isn’t set too aggressively, and verify that the original image wasn’t already clipped.
  • Color shift after correction → ensure “Preserve color balance” is enabled; otherwise, consider a color calibration step afterwards.
  • Residual gradient remains → try a smaller tile size or a higher order/smoothing setting and re-run the model.

Updated 2025-11-15