Match Monochrome Channel Masters with LinearFit
Choose a reference, test intensity matching on linear monochrome masters, and inspect the fit before RGB, SHO, or HOO combination. Learn when matching is useful, when to skip it, and why it does not replace color calibration or better data.
Sources reviewed October 25, 2026
What LinearFit changes
LinearFit matches a target image’s intensities to a reference using a fitted straight-line relationship: a scale and an offset. It uses corresponding pixel locations, so the images must already be registered. It does not align stars or remove a spatially varying gradient.
Use it as an optional comparison before combining linear monochrome masters. For this exercise, complete calibration, registration, integration, a common crop, and appropriate gradient correction first. Keep stars present and do not permanently stretch the inputs.
Intensity matching is not physical color calibration. It also does not create signal, improve the signal-to-noise ratio merely by rescaling values, or preserve original emission-line ratios. Retain an unfitted baseline and decide whether the fitted combination helps your intended presentation.
1. Prepare comparable masters
Open the gradient-corrected R/G/B masters, or S II/Hα/O III for SHO, or Hα/O III for HOO. Confirm identical dimensions, orientation, common crop, and pixel registration by comparing stars at 1:1. Matching dimensions alone is insufficient.
Save untouched checkpoints, then duplicate the working views with clear identifiers such as Target_R_fit, Target_G_reference, and Target_B_fit. For a narrowband exercise, preserve the actual filter identity in each name. A later display-channel assignment does not change what was acquired.
Check for uncovered edges, severe clipping, residual gradients, and calibration artifacts before fitting. A global fit cannot solve these. Keep the same geometry through the experiment and return to a fresh duplicate whenever comparing a new configuration.
2. Choose one sound reference
Choose a clean master with useful unsaturated signal, acceptable noise, and a sound background correction. Avoid choosing solely by the brightest STF display, darkest median, or smallest numerical noise value: scaling and display settings affect those comparisons.
For an introductory RGB trial, a good G master can be a convenient reference if its quality supports that choice; R or B can also be appropriate. Leave the chosen reference unchanged and fit the other two directly to it. Do not fit R to G and then use the altered R as a new reference for B.
For SHO or HOO, a good Hα master can be one trial reference, but it is not a universal prescription. Different emission lines can trace different structures, so their corresponding pixels may not follow one useful straight-line relationship. If matching produces implausible results, retain the unfitted channels and handle the display palette later.
This lesson matches the color-channel masters to one of their own set. It does not use a separate luminance master as a compulsory reference or perform an LRGB blend.
3. Run a controlled fit
Find LinearFit through All Processes, usually under ColorCalibration. Use the installed process help for current control definitions; the linked examples use older interfaces.
- Reset a reused instance. Select the unchanged reference view explicitly in Reference image. Check the view identifier rather than assuming the active image is the reference.
- Review Reject low and Reject high. These define which pixel values participate in estimating the fit; they are not new black and white points or star-rejection sigma settings. Start with the installed defaults and verify that enough valid, non-clipped data remains in both images. Do not copy arbitrary limits from another target.
- For the RGB example with G as reference, apply LinearFit to the R working duplicate by dragging its process-instance triangle onto that target view. Read the console for success and any fit information or warnings.
- Without changing or modifying the G reference, apply the same configuration to the B working duplicate. Do not apply it to the reference itself.
- Save successful targets separately, for example Target_R_LinearFit_to_G.xisf and Target_B_LinearFit_to_G.xisf. Save the reference and process instance with the experiment. If you alter the rejection limits, rerun from fresh target copies rather than stacking fits.
4. Inspect the targets before combination
Compare each original and fitted target with the same STF, framing, and zoom. An independently calculated auto-stretch on each view can hide the intended intensity change. You can also inspect each result with a fresh STF afterward, but distinguish that display change from the pixel transformation.
Inspect bright cores, faint structures, smooth sky, and borders. Look for new clipping, extreme offsets, or a noise-dominated channel being amplified to an unusable level. If reported fit coefficients are surprising, check the reference, registration, gradients, and retained pixel range before accepting them.
A lower numerical noise value after rescaling does not prove improved data quality. Likewise, similar-looking histograms do not prove that astrophysical structures were matched appropriately. Compare the final fitted and unfitted combinations as well as the individual channels.
5. Apply the appropriate channel path
Always preserve a combination of the original corrected masters as your baseline. Use the same display treatment when comparing it with the fitted trial.
| Path | Example fit trial | ChannelCombination assignment |
|---|---|---|
| RGB | Keep a suitable G reference; fit R and B directly to it | R → red, G → green, B → blue |
| SHO | Keep a suitable Hα reference; test S II and O III fits separately | S II → red, Hα → green, O III → blue |
| HOO | Keep a suitable Hα reference; test O III once | Hα → red; the same accepted O III → green and blue |
- Open ChannelCombination, choose RGB, and select the intended reference and accepted fitted targets. Verify actual filter identities against the assignments above.
- Use Apply Global to create a new linear composite. For HOO, reuse the same accepted O III view in green and blue; do not make different fitted versions simply because the display has two slots.
- Inspect with a linked RGB STF and compare with the unfitted baseline. Judge gradients, color balance, faint structures, noise, and star profiles together. A more neutral first display is not proof of physically correct color.
- For broadband RGB, continue to the appropriate sky-solution checks and SPCC calibration. For SHO/HOO, continue to the narrowband palette workflow. NarrowbandNormalization later in that workflow is a separate display-balancing operation; it does not make LinearFit mandatory.
When to skip or reject the fit
Skip it when the existing combination and subsequent calibration already meet the purpose, or when preserving the acquired relative intensities is essential. LinearFit is a processing choice, not a prerequisite for ChannelCombination or SPCC.
Be particularly cautious with narrowband masters whose emission structures differ strongly. A successful numerical fit can still be a poor representation of the field. Never interpret an aesthetically balanced SHO image as a measurement of elemental abundance or physical line ratios.
This exercise does not require splitting an existing one-shot-color RGB integration just to add another processing step. It also does not replace WBPP normalization, MGC’s SPFC prerequisite, or calibration-frame matching.
If returning to MGC after a fit, use a suitable preserved linear checkpoint and regenerate flux-calibration prerequisites as needed. Do not assume metadata describing the earlier pixel values remains valid after intensity transformations.
Troubleshooting and saving the decision
Fit fails or retains too little data: confirm geometry, image state, rejection limits, and the presence of usable corresponding pixels. Do not widen limits blindly to include saturated cores or invalid borders.
Unexpected color or dominant noise: inspect the reference and each target independently, then compare the unfitted combination. Scaling a weak channel cannot replace additional exposure time.
Stars have colored fringes: check registration and channel star profiles. An intensity fit cannot repair shifted or differently blurred stars.
A second trial is hard to interpret: return to the same untouched targets, keep the reference fixed, and change one setting. Do not combine a new fit with simultaneous stretching or palette adjustments.
Keep the corrected originals, unchanged reference, accepted fitted targets, both combination baselines, and the process instance. Record the reference identifier, rejection limits, target list, and why you kept or skipped the operation.
This is an original practice sequence. The Jon Rista reference explains matching and examples; the Adam Block page lists further training, whose paid content is not reproduced here. Installed PixInsight documentation governs version-specific controls.
See the technique in practice
This lesson is public. Detailed project workflows and dataset downloads require an active membership. Dataset selections follow your membership allowance; reading a workflow does not use a selection.
Let the shell stand alone
See this technique in the recorded Crescent Nebula - NGC6888 workflow. Use that project’s documented order, retained settings, and limitations.
Public overview · Member-only detailed workflow
Explore story and member workflow →Finding the Pelican in blue and gold
See this technique in the recorded Pelican Nebula - IC5070 workflow. Use that project’s documented order, retained settings, and limitations.
Public overview · Member-only detailed workflow
Explore story and member workflow →