Reveal Bright-Core Detail with HDRMultiscaleTransform
Control the brightness range of a stretched nebula or galaxy core with HDRMultiscaleTransform. Compare layer scales, use lightness and region masks, and preserve a convincing transition from the core to its surroundings.
Sources reviewed October 25, 2026
What HDRMT can reveal
A bright nebular center or galaxy core can dominate a stretched image even when useful structure remains in its pixels. HDRMultiscaleTransform, often called HDRMT, compresses brightness differences at selected spatial scales so that structure can become easier to see alongside fainter surroundings.
You may see this technique called HDRMultiscaleTransformation, but the native PixInsight process is HDRMultiscaleTransform. Find it through All Processes. This exercise uses it after a permanent stretch as an optional finishing step, before final local contrast and star restoration in a starless workflow.
HDRMT does not reconstruct saturated pixels. If the original exposures clipped the core, shorter exposures and a separate HDR-combination workflow may be needed. If clipping happened only during processing, return to a preserved linear checkpoint and make a gentler stretch first.
1. Check the source and save a baseline
Begin with a calibrated, gradient-corrected image that has received its intended color preparation and permanent stretch. Disable STF: the image should remain visibly stretched. Save Target_before_HDRMT.xisf and create a working duplicate.
Inspect the core with pixel readouts and the histogram as well as by eye. A flat region pinned at the maximum value may be clipped rather than merely bright. Check the linear source if uncertain; do not assume HDRMT will reveal texture where the data contains none.
For a separated-star workflow, use the stretched starless object and keep the prepared stars separate. Star-removal artifacts can also be intensified, so inspect the core and nearby bright-star positions before starting. With stars intact, expect to evaluate their protection carefully.
Set aside comparison areas containing the core, its surrounding transition, and faint outer structure. Preview regions should include substantial context; a tiny core-only crop can hide a dark ring or an unnatural brightness relationship.
2. Understand scale and protection
Reset a reused HDRMT instance and consult the installed help for the current controls. Keep advanced options fixed while comparing the main scale setting.
| Control | Role in the trial | What to check |
|---|---|---|
| Number of layers | Selects the multiscale size of the treatment | Compare nearby layer counts; lower counts can make finer-scale changes conspicuous |
| Iterations | Repeats the transformation | Start with one; repeated application can flatten structure |
| To lightness | On color data, directs the operation to lightness rather than treating RGB channels independently | It is not a protective region mask; compare color appearance afterward |
| Lightness mask | Built-in protection that favors brighter regions | It does not identify the desired core or automatically protect every star |
| External image mask | Lets you choose where the treatment is permitted | Check geometry, inversion, and smooth transitions |
3. Build a simple core-region mask when needed
For localized control, extract CIE L* from the stretched color image using ChannelExtraction in CIE L*a*b* mode, or duplicate a grayscale target. Keep the resulting mask aligned and at the same dimensions as the working image.
On the mask copy, use HistogramTransformation or CurvesTransformation to darken the background and faint outskirts while leaving the intended bright region lighter. Keep a broad, gradual transition around the core. A hard white island on black can create a visible seam in the result.
Select the mask with Mask → Select Mask and enable it on the HDRMT target. White permits the effect, black protects, and gray permits a partial effect. Inspect the mask itself and its overlay, verifying whether inversion is enabled.
A lightness-derived mask also includes bright stars. Use the starless branch for this first exercise when available, or add an appropriate star-protection strategy before proceeding on stars-intact data. The external mask and built-in Lightness mask provide different forms of protection and can work together.
4. Compare a one-pass trial
These settings are a practice comparison, not a universal core-recovery preset. The useful scale depends on image sampling and the size of the bright region.
- On a fresh working copy, start with Number of layers 6 and Iterations 1. For a color image, enable To lightness; it is not needed for an already grayscale target.
- Enable the built-in Lightness mask for this baseline. Verify the external core mask, if used, is enabled and protecting the intended surroundings. Leave other algorithm options at their installed defaults and record them.
- Apply once to a representative preview or disposable full-image duplicate. Keep STF off. Inspect the core and the transition into the surrounding object rather than judging the core alone.
- Return to the same untouched baseline and compare a second trial at 7 or 8 layers with all other settings unchanged. Do not apply each new trial on top of the preceding one.
- If every trial is too strong, reconsider the scale and reduce the permitted effect through a softer or darker external mask. Do not look for LHE’s Amount control in HDRMT; they are different processes.
5. Keep the core believable
Compare source and candidate at the same zoom and display conditions, then at the intended viewing size. The core may remain the brightest part of the object while showing more internal structure. Making its brightness equal to the outskirts is not the goal.
Look for gray or flattened cores, dark bowls, bright rims, harsh texture, altered star profiles, and abrupt mask boundaries. Inspect faint surroundings for unintended changes too. A newly visible pattern can be amplified noise or an earlier artifact rather than additional physical detail.
Check color and saturation as well as lightness, even with To lightness enabled. If the original brightness hierarchy or attractive core glow is lost, prefer a gentler trial or skip the operation.
A preview result is provisional. Validate on a fresh complete image because the surrounding context and multiscale neighborhoods differ from those in a small preview. Keep both baseline and candidate for an honest comparison.
6. Continue with restrained finishing
Apply the accepted configuration once to a fresh full-image copy and inspect again. Save Target_after_HDRMT.xisf with the pre-HDRMT image, process instance, and any mask. Record the layer count, iterations, lightness options, other retained defaults, and the reason for accepting the result.
If desired, test LHE afterward on a duplicate for local contrast. HDRMT manages brightness range while LHE enhances local contrast; neither is mandatory, and their effects can accumulate. Compare the combined result with the original baseline, not only with the preceding intermediate image.
For a starless workflow, restore the separately processed stars after accepting the object’s core and contrast. Inspect their surroundings in the new composite. Finish selective color work as appropriate, then use the export lesson to create web and print copies.
Quick troubleshooting
No core detail appears: verify that structure exists in the source, the image is permanently stretched, and the mask allows the operation. Clipped data cannot be recovered by increasing iterations.
The center becomes dark or gray: return to the baseline, compare a different layer scale, and reduce the effect allowed by the external mask. Do not compensate by applying an aggressive brightness curve over an already flattened result.
Rings or halos appear: inspect mask transitions, scale choice, nearby bright stars, and source artifacts. Review any deringing controls in the installed help only after identifying the cause; they are not a guarantee that a strong trial is acceptable.
The result is crunchy or noisy: simplify to one pass and inspect whether the selected scale is amplifying weak structures or noise. Additional denoising is not automatically the right repair.
Only the small preview looks good: compare the whole object and brightness transition. Accept a full-image result only after inspecting it at 1:1 and presentation size.
This is an original beginner practice sequence. The linked material provides masking concepts and worked examples with older interfaces; installed process documentation governs current behavior. The lesson does not claim a hands-on test of every control in the latest build.