Narrowband workflow (HOO/SHO)
Processing narrowband data from a monochrome camera – Ha and OIII for HOO, plus SII for SHO – from stacking to export. The stars come either from the narrowband image or from short RGB exposures.
Overview
| Decision | Variant A | Variant B |
|---|---|---|
| Palette | HOO: Ha red, OIII teal – close to natural | SHO: SII red, Ha green, OIII blue – Hubble palette |
| Stars | from the narrowband image – no further data needed, star colours meaningless | from RGB – natural star colours, needs short RGB exposures |
The variants appear as tabs where they matter. Image names in the text: Ha, OIII, SII, the combined colour image HOO or SHO, the luminance L_syn, and RGB for stars from RGB.
1 Calibrate and integrateLinear
Goal: aligned masters for each filter.
ToolsChains calibration, cosmetic correction, weighting, registration, normalisation and integration into a reproducible pipeline – from individual subs to aligned master lights. Open fundamentals →
One flat and one light group per filter, CFA off (Chains calibration, cosmetic correction, weighting, registration, normalisation and integration into a reproducible pipeline – from individual subs to aligned master lights. Open fundamentals →). Important for narrowband:
- One common registration reference for all filters, ideally the best Ha frame (many well-detected stars). With stars from RGB, include the RGB lights in the same WBPP run.
- Drizzle, if at all, the same for all filters.
- Enable Generate rejection maps.
2 CropLinear
Goal: remove stacking borders and stray-light streaks before gradient correction mistakes them for gradients.
Set DynamicCrop on one master and apply the same crop to all others (drag the process icon onto each image) so that they stay aligned. Artefacts missing from the rejection maps were present in all subs (Chains calibration, cosmetic correction, weighting, registration, normalisation and integration into a reproducible pipeline – from individual subs to aligned master lights. Open fundamentals →) – crop them away too.
3 Gradient correction per channelLinear
Goal: remove background gradients while each channel is still separate.
ToolsModels large-scale brightness and colour gradients (light pollution, moon, residual vignetting) and subtracts them – structures are kept out of the model by protection masks. Open fundamentals →
Whyper channel and before combining?
Gradients differ per filter; moonlight is stronger in blue, light pollution is not spectrally neutral. After combining, they would be merged into colour gradients and harder to separate.
Correct each master separately with Models large-scale brightness and colour gradients (light pollution, moon, residual vignetting) and subtracts them – structures are kept out of the model by protection masks. Open fundamentals →. In narrowband, emission often fills large parts of the image: keep the Protection Threshold low (Models large-scale brightness and colour gradients (light pollution, moon, residual vignetting) and subtracts them – structures are kept out of the model by protection masks. Open fundamentals →). Be especially careful with weak OIII and SII – the structure lies only just above the background.
MultiscaleGradientCorrection is not intended for pure narrowband masters (MARS is based on broadband data); it can be used for R, G, B.
4 RGB for the stars (optional)Linear
Goal: only for stars from RGB – a colour-calibrated image that will later supply only the stars.
ToolsCalibrates the colours of a broadband or OSC image by comparing thousands of stars with Gaia spectra – objective and reproducible, including background neutralisation. Open fundamentals →
- ChannelCombination:
R,G,B→RGB. - ImageSolver, if the solution does not come from WBPP.
- Calibrates the colours of a broadband or OSC image by comparing thousands of stars with Gaia spectra – objective and reproducible, including background neutralisation. Open fundamentals → with sensor and RGB filter curves, region of interest on a nebula-free preview.
WhySPCC even for a star-only image?
The star colours are the RGB’s only contribution to the final image. Only a photometrically correct white balance gives the natural range from orange to blue. The nebula colour in the RGB does not matter – it is discarded with the star separation.
5 DeconvolutionLinear
Goal: correct star shapes and sharpen structures while the data are linear.
ToolsAI-based deconvolution for linear data – corrects star shapes, makes stars smaller and sharpens nebula structure. Open fundamentals →
AI-based deconvolution for linear data – corrects star shapes, makes stars smaller and sharpens nebula structure. Open fundamentals → on each master, first Correct Only, then normal mode. Keep Sharpen Stars and Adjust Star Halos the same in all channels: Ha stars are narrow, OIII stars often larger; different profiles later produce coloured fringes. Sharpen Nonstellar may differ (Ha takes more than OIII and SII).
6 Linear denoisingLinear
Goal: remove noise while it is evenly distributed – above all in the weak channels.
ToolsAI-based noise reduction with separate strength for fine- and large-scale noise and for brightness and colour. Several light passes keep more detail than one strong pass. Open fundamentals →
Whynow?
In the linear image noise is evenly distributed and easy to separate from the signal. OIII and SII must later be stretched more than Ha; their noise would then become disproportionately visible.
AI-based noise reduction with separate strength for fine- and large-scale noise and for brightness and colour. Several light passes keep more detail than one strong pass. Open fundamentals →: Ha HF 0.4–0.6 / LF 0.2–0.3; OIII and SII HF 0.6–0.8 / LF 0.3–0.4. RGB (if present): AI-based noise reduction with separate strength for fine- and large-scale noise and for brightness and colour. Several light passes keep more detail than one strong pass. Open fundamentals →.
7 Synthetic luminanceLinear
Goal: a luminance that carries each structure with its strongest signal.
ToolsA luminance built from existing channels, so that structure and colour can be processed separately and merged again with LRGBCombination. Open fundamentals →Combines images pixel by pixel with formulas – for combining channels, blending luminances or recombining stars. Open fundamentals →
Whynot simply Ha as luminance?
Ha is usually the strongest channel. But a luminance from Ha alone would suppress structures that shine mainly in OIII or SII – they would end up dark in the final image although the colour image contains them.
- Match copies of
OIII(andSII) toHawith A luminance built from existing channels, so that structure and colour can be processed separately and merged again with LRGBCombination. Open fundamentals → →OIII_LF,SII_LF. - A luminance built from existing channels, so that structure and colour can be processed separately and merged again with LRGBCombination. Open fundamentals →:
max(Ha, OIII_LF)ormax(Ha, OIII_LF, SII_LF); with noisy OIII weighted, e.g.Ha*0.6 + OIII_LF*0.4.
8 Stretching and colour combinationNon-linear
Goal: make all channels visible, bring them to comparable levels and only then combine them into the colour image.
ToolsStretches linear data so that contrast is created around a chosen brightness value – in several small passes, alternating with a black point. Open fundamentals →Combines images pixel by pixel with formulas – for combining channels, blending luminances or recombining stars. Open fundamentals →Gleicht die Kanäle so an, dass der Hintergrund neutral grau wird – ohne die Farben des Objekts zu kalibrieren. Für Falschfarben und als Nachkorrektur. Open fundamentals →Measures at fixed spots whether a separate luminance and the colour image match in background, midtones and highlights, and adjusts the luminance with a curve – so the LRGB combination is neither pale nor oversaturated. Open fundamentals →
Whycombine non-linearly?
The channels often differ in signal strength by orders of magnitude (Ha 5–10× stronger than OIII and SII). Combined linearly, Ha would dominate the colour image completely. Stretched individually and brought to similar levels, all channels contribute equally to colour – the colour balance becomes something you shape.
8.1 Stretching and matching the channels
Stretch each channel and L_syn with Stretches linear data so that contrast is created around a chosen brightness value – in several small passes, alternating with a black point. Open fundamentals →. Then match the channels:
| Quantity | Target | Why |
|---|---|---|
| background (median of a background preview) | nearly equal in all channels, about 0.08–0.12 | unequal backgrounds give a colour cast |
| brightness of structures | similar; do not let weak channels sink | a channel that is too dark is almost absent from the colour image |
Create previews Dark and Bright in Ha, transfer them to the other channels, read the median with Statistics; match deviations with a black point (background) or a Remaps tonal values along a freely shaped curve – for contrast, saturation, hue and for shifting individual tonal ranges. Acts globally; locally only with a mask. Open fundamentals → (structure brightness). OIII and SII usually need more stretching; if noise becomes visible, a light NXT pass (0.3–0.4).
Linear alternative: LinearFit of the weaker channels to Ha, ChannelCombination, joint stretch in Colour mode – the result is dominated by Ha, weak channels recede; for a natural ratio of the emission lines.
8.2 Combination
Combines images pixel by pixel with formulas – for combining channels, blending luminances or recombining stars. Open fundamentals →, Create new image, RGB, ID HOO: R = Ha, G = OIII, B = OIII. Variant with more natural transitions: G = Ha*0.3 + OIII*0.7.
Result: Ha red, OIII teal, mixed areas whitish.
ChannelCombination (RGB): R = SII, G = Ha, B = OIII → SHO.
Expected result: strongly greenish, because Ha sits in G and is usually strongest. Not an error – the palette is created in step 10.
8.3 Neutralising the background
Gleicht die Kanäle so an, dass der Hintergrund neutral grau wird – ohne die Farben des Objekts zu kalibrieren. Für Falschfarben und als Nachkorrektur. Open fundamentals → with a nebula-free preview as reference. No SPCC for false colour – there is no “correct” star colour; only the background should be neutral.
8.4 Matching luminance to the colour image
Before separating the stars, match L_syn to the colour image’s luminance (Measures at fixed spots whether a separate luminance and the colour image match in background, midtones and highlights, and adjusts the luminance with a curve – so the LRGB combination is neither pale nor oversaturated. Open fundamentals →): ChannelExtraction on HOO or SHO → HOO_L or SHO_L. Besides Dark, Mid, Bright, add a measuring point OIII on dominant OIII emission. Typically only the midtones need adjusting, with a Measures at fixed spots whether a separate luminance and the colour image match in background, midtones and highlights, and adjusts the luminance with a curve – so the LRGB combination is neither pale nor oversaturated. Open fundamentals →, because L_syn deliberately emphasises Ha structure.
Whynow, before star separation?
Stars lie almost entirely above the fixed bright point and remain unchanged. Stars and structure then relate the same way in both images – the separated images will fit together later.
9 Separating the starsNon-linear
Goal: be able to process nebula and stars separately.
ToolsSplits an image into a starless image and a star image, so that nebula and stars can be processed separately and recombined at the end with a screen blend. Open fundamentals →
Whyafter stretching?
StarXTerminator works most reliably on non-linear data; halos and faint stars are then fully visible.
| Image | Generate star image | Unscreen stars |
|---|---|---|
HOO / SHO | enabled | enabled |
L_syn | off | – |
The stars come from the colour image; without RGB data it is the only coloured source (Splits an image into a starless image and a star image, so that nebula and stars can be processed separately and recombined at the end with a screen blend. Open fundamentals →).
| Image | Generate star image | Unscreen stars |
|---|---|---|
HOO / SHO | off | – |
L_syn | off | – |
RGB | enabled | enabled |
The starless RGB is discarded. Because all masters were registered to the same reference and cropped identically, RGB_stars is aligned.
Enable Large overlap for dense star fields.
10 Processing the starless imagesNon-linear
Goal: colour and colour contrast in the colour image, structure in the luminance – without the stars reacting.
ToolsRemoves an excess of one colour, usually green, by limiting that channel to the level of the other two – a correction of colour errors or a deliberate creative tool. Open fundamentals →Remaps tonal values along a freely shaped curve – for contrast, saturation, hue and for shifting individual tonal ranges. Acts globally; locally only with a mask. Open fundamentals →AI-based noise reduction with separate strength for fine- and large-scale noise and for brightness and colour. Several light passes keep more detail than one strong pass. Open fundamentals →Raises local contrast – structures that differ only slightly from their immediate surroundings stand out without shifting the global tonal values. Open fundamentals →Creates a mask from a brightness range – for example "everything brighter than the background" – so that a following process acts only there. Open fundamentals →
Whynow and in this order?
Without stars, contrast and saturation steps act only on the nebula. In the colour image the colour is cleaned up or the palette shaped first, then contrast is built, then saturation raised and finally the image denoised – otherwise colour errors would be amplified, and noise removed before a contrast boost would become visible again.
10.1 Colour image
1. Colour:
No blanket SCNR. OIII lies in G and B and appears teal; SCNR would shift it towards blue and desaturate it. Only for a real green cast in the background use Removes an excess of one colour, usually green, by limiting that channel to the level of the other two – a correction of colour errors or a deliberate creative tool. Open fundamentals →, Amount 0.3–0.5; compare before/after at an OIII spot.
Optional – red/teal separation:
| Goal | Tool | Procedure |
|---|---|---|
| OIII a stronger teal | ColorSaturation | point in the cyan/blue-green range +0.1 to +0.2 |
| Ha warmer | Remaps tonal values along a freely shaped curve – for contrast, saturation, hue and for shifting individual tonal ranges. Acts globally; locally only with a mask. Open fundamentals → | shift reds slightly towards orange |
| Ha/OIII transitions | – | keep whitish mixed areas – they show the overlap of both emissions |
Palette with SCNR: Removes an excess of one colour, usually green, by limiting that channel to the level of the other two – a correction of colour errors or a deliberate creative tool. Open fundamentals →, Average or Maximum Neutral, Amount 0.5–1.0, preferably several passes. Areas with SII turn gold, with OIII blue.
Shift hues with Remaps tonal values along a freely shaped curve – for contrast, saturation, hue and for shifting individual tonal ranges. Acts globally; locally only with a mask. Open fundamentals →:
| Goal | Procedure |
|---|---|
| yellow → gold/orange | point in the yellow range slightly towards red |
| green residue → yellow | point in the green range towards yellow |
| blue → cyan/teal | point in the blue range towards cyan |
Small shifts only; then check the background for neutrality again.
2. Contrast: Remaps tonal values along a freely shaped curve – for contrast, saturation, hue and for shifting individual tonal ranges. Acts globally; locally only with a mask. Open fundamentals →, moderate.
3. Saturation: Remaps tonal values along a freely shaped curve – for contrast, saturation, hue and for shifting individual tonal ranges. Acts globally; locally only with a mask. Open fundamentals →. Goal: palette colours stronger, background neutral.
4. Denoising: AI-based noise reduction with separate strength for fine- and large-scale noise and for brightness and colour. Several light passes keep more detail than one strong pass. Open fundamentals → – in the LRGB combination the luminance replaces brightness; colour noise remains visible.
10.2 Luminance
Order: local contrast → optionally midtones with a mask → denoising.
WhyLHE before denoising?
LHE also makes noise visible. Denoising comes afterwards so that exactly this noise is captured.
- Raises local contrast – structures that differ only slightly from their immediate surroundings stand out without shifting the global tonal values. Open fundamentals →: 64–128 for pillars and edges, 16–32 for fine structure.
- Optional: lift faint nebula areas with a Creates a mask from a brightness range – for example "everything brighter than the background" – so that a following process acts only there. Open fundamentals →; remove the mask afterwards.
- AI-based noise reduction with separate strength for fine- and large-scale noise and for brightness and colour. Several light passes keep more detail than one strong pass. Open fundamentals →.
11 Combining and backgroundNon-linear
Goal: merge structure and colour, tune the background.
ToolsMeasures at fixed spots whether a separate luminance and the colour image match in background, midtones and highlights, and adjusts the luminance with a curve – so the LRGB combination is neither pale nor oversaturated. Open fundamentals →Replaces the brightness of a colour image with a separate luminance – structure and sharpness from the luminance, colour from the colour image. Open fundamentals →Remaps tonal values along a freely shaped curve – for contrast, saturation, hue and for shifting individual tonal ranges. Acts globally; locally only with a mask. Open fundamentals →Gleicht die Kanäle so an, dass der Hintergrund neutral grau wird – ohne die Farben des Objekts zu kalibrieren. Für Falschfarben und als Nachkorrektur. Open fundamentals →
Whymatch again?
Since the first matching, both images have changed differently: the colour image gained contrast through the S-curve and saturation, the luminance changed only locally through LHE. The luminance now typically has less global contrast; without matching, the combination looks flat.
- Matching: ChannelExtraction on
HOO_starlessorSHO_starless→…_starless_L; measuring points as in 8.4, includingOIII. Now typically a Measures at fixed spots whether a separate luminance and the colour image match in background, midtones and highlights, and adjusts the luminance with a curve – so the LRGB combination is neither pale nor oversaturated. Open fundamentals →. - Combination: Replaces the brightness of a colour image with a separate luminance – structure and sharpness from the luminance, colour from the colour image. Open fundamentals → with
L_syn_starlessas L on the colour image; Lightness 0.5, Saturation 0.4 to start, find values on duplicates. For loss of detail see the Replaces the brightness of a colour image with a separate luminance – structure and sharpness from the luminance, colour from the colour image. Open fundamentals →. - Darken the background if needed: Remaps tonal values along a freely shaped curve – for contrast, saturation, hue and for shifting individual tonal ranges. Acts globally; locally only with a mask. Open fundamentals →.
- Check neutrality: Gleicht die Kanäle so an, dass der Hintergrund neutral grau wird – ohne die Farben des Objekts zu kalibrieren. Für Falschfarben und als Nachkorrektur. Open fundamentals → – near strong colours, simultaneous contrast deceives.
12 Processing the starsNon-linear
Goal: stars that suit the nebula and do not cover it.
ToolsRemaps tonal values along a freely shaped curve – for contrast, saturation, hue and for shifting individual tonal ranges. Acts globally; locally only with a mask. Open fundamentals →Removes an excess of one colour, usually green, by limiting that channel to the level of the other two – a correction of colour errors or a deliberate creative tool. Open fundamentals →Makes stars smaller instead of merely dimming them – faint stars recede, bright ones remain as accents. Easiest in a separated star image. Open fundamentals →
On a copy of the star image – it contains only stars on a black background, so processes can be applied without a star mask.
1. Colour:
Whyshape rather than calibrate?
Narrowband stars show no real star colours: Ha-strong stars appear reddish, OIII-strong ones cyan – depending on spectrum and filter profiles, not temperature. They are therefore reduced for visual effect.
| Variant | Procedure | Effect |
|---|---|---|
| subtly coloured | lower Remaps tonal values along a freely shaped curve – for contrast, saturation, hue and for shifting individual tonal ranges. Acts globally; locally only with a mask. Open fundamentals → slightly (0.5 → 0.4) | light palette tones |
| neutral | lower channel S strongly or ColorSaturation negative | white stars that do not distract from the nebula |
| magenta fringes (SHO, HOO with G blend) | Removes an excess of one colour, usually green, by limiting that channel to the level of the other two – a correction of colour errors or a deliberate creative tool. Open fundamentals → | fringes neutral |
- Match size and brightness: the stars come from other exposures (different seeing, different exposure). Too large or bright → stronger reduction,
s< 1 when recombining; too small or faint → stretch RGB a little more ors= 1.0. - Greenish stars: Removes an excess of one colour, usually green, by limiting that channel to the level of the other two – a correction of colour errors or a deliberate creative tool. Open fundamentals →, Amount 0.5–0.8.
- Check saturation: star colours strong but not garish; too colourful after recombination → lower channel S slightly. Do not remove completely – the colours show the temperature distribution.
2. Star reduction: Makes stars smaller instead of merely dimming them – faint stars recede, bright ones remain as accents. Easiest in a separated star image. Open fundamentals →, Selection 0.2–0.3, Amount 0.5–0.8, 5×5 circular.
13 Recombining the starsNon-linear
Goal: put the stars back softly and without overshoot.
ToolsSplits an image into a starless image and a star image, so that nebula and stars can be processed separately and recombined at the end with a screen blend. Open fundamentals →Combines images pixel by pixel with formulas – for combining channels, blending luminances or recombining stars. Open fundamentals →
Splits an image into a starless image and a star image, so that nebula and stars can be processed separately and recombined at the end with a screen blend. Open fundamentals → with PixelMath, Replace target image, on a duplicate of Starless_final:
~((~Starless_final) * (~(HOO_stars * s)))Replace HOO_stars with SHO_stars or RGB_stars. s = 1.0 after reduction; 0.8–0.9 if RGB stars look too dominant.
14 Finishing and exportNon-linear
Goal: overall check, last small corrections, saving.
ToolsRemaps tonal values along a freely shaped curve – for contrast, saturation, hue and for shifting individual tonal ranges. Acts globally; locally only with a mask. Open fundamentals →AI-based noise reduction with separate strength for fine- and large-scale noise and for brightness and colour. Several light passes keep more detail than one strong pass. Open fundamentals →
- Optional curve: small corrections of brightness and contrast only – with stars in the image, strong curves bloat them again.
- Optional NXT: only for visible noise in the background, AI-based noise reduction with separate strength for fine- and large-scale noise and for brightness and colour. Several light passes keep more detail than one strong pass. Open fundamentals → (HF Intensity 0.15–0.25, LF 0.1). No visible difference → skip it.
| Export | Purpose |
|---|---|
| XISF, 32-bit float | archive – lossless, with history and metadata |
| TIFF 16-bit | further editing in other software |
| JPEG, sRGB, downsized if needed | publication |
Name and arrange the process icons in workflow order, Save Selected Icons…. Save the project after the last change; the message “images have been modified and not saved” on closing refers to individual files, not to the project.
Common problems
Narrowband-specific. General errors are listed under “Typical errors” of each fundamentals article.
| Symptom | Cause | Remedy |
|---|---|---|
| SHO image strongly green | Ha in G dominates | expected; palette with SCNR in step 10 |
| Nebula red or teal on one side, one channel dominates | channels stretched differently before combining | repeat channel matching 8.1 |
| OIII or SII structures dark in the final image | luminance from Ha only | L_syn from all channels (step 7) |
| OIII areas blue or grey (HOO) | SCNR on the nebula | SCNR only masked in the background |
| Noise in OIII/SII after stretching | weak signal | stronger linear NXT; light pass after stretching |
| Stars red/cyan, restless | false colours of narrowband stars | desaturate stars or use stars from RGB |
| Magenta stars, coloured rims | different star profiles in the channels | same BXT star parameters; Invert → SCNR → Invert |
| Double stars or offset after recombination (RGB) | RGB not registered to the same reference | StarAlignment of RGB to Ha, crop again |
| RGB stars too large | stronger stretch, different seeing | stronger reduction, s < 1 |
| Unnatural star colours (RGB) | SPCC inaccurate or RGB stretched in RGB mode | check the SPCC fits; GHS mode Colour |