Stretch with GHS: When to Choose It or MAS

A permanent stretch turns a dark linear image into a display-ready image. GeneralizedHyperbolicStretch, or GHS, gives you direct control over the shape and focus of an intensity transformation. MultiscaleAdaptiveStretch, or MAS, offers an adaptive multiscale approach. Both belong at the stretching stage, after the appropriate linear preparation.

Use these as alternative first-stretch experiments on separate copies. You do not need to apply MAS and then GHS as a required sequence. Neither can recover clipped data or turn weak signal into additional exposure time.

The comparison below is practical guidance, not a claim that either tool always produces a better image. A bright galaxy core, faint dust, and a starless narrowband nebula can call for different compromises.

Use the GHS publisher’s current installation instructions and compatibility information. Its process module is GeneralizedHyperbolicStretch; older tutorials may use a script with different spelling and controls. The publisher states that from PixInsight 1.9.4 its repository installs the process rather than the superseded script.

Begin with an integrated, gradient-corrected linear image after the appropriate channel combination and calibration. Complete operations that require linear data before stretching. Save Target_linear_ready.xisf, then make separate Target_GHS_trial and Target_MAS_trial copies.

If using StarXTerminator, follow the star-separation lesson and extract stars before this stretch. GHS can change star profiles in ways that complicate later removal. Keep a separate stars branch and use the same star-restoration setup when comparing finished object branches.

Find GeneralizedHyperbolicStretch through All Processes and reset the instance. Confirm the intended view is active. For the first global-stretch exercise, disable unintended masks and record that choice.

Keep a small set of controls in focus and use the installed help for version-specific labels. These controls describe intensity ranges, not particular physical objects: a star and a nebular feature can share similar pixel values.

Start with a conservative global RGB/K trial. This simplifies the color options for learning; more specialized color-preserving modes can be compared later. Do not adjust red, green, and blue independently to conceal an unfinished calibration problem.

A first stretch need not finish every brightness range. From a saved pass-one duplicate, identify a midtone region that looks flat. Use its current pixel values for a new SP trial; the original linear readout no longer describes this stretched image.

Begin the second trial at zero Stretch factor again so a saved first-pass instance is not accidentally repeated. Increase gently and compare broader versus more concentrated Local intensity. Keep the first-pass result available and save the new configuration separately.

Explore shadow and highlight protection only to address an observed problem. They reshape the distribution of contrast; they are not absolute locks that guarantee unchanged background or star cores. Recheck the rest of the image whenever changing protection.

If a trial loses faint dust, weakens a bright core excessively, or amplifies noise, return to the prior checkpoint. Do not keep adding passes until every region looks equally prominent. Record the sequence of accepted passes so the result can be reproduced.

For a later comparison, test GHS’s Colour mode on a fresh copy at the same starting stage. Read the documentation for Colour blend and Clip type before using them. The mode can retain stronger color, but bright channel values need appropriate handling; it is not an automatic guarantee against clipping.

Inspect bright stars, saturated nebular regions, and smooth color transitions. Choose the mode and strength on the basis of the image, not the most vivid preview. Save the mode and any clip-handling choices with each pass.

Keep highlight recovery expectations realistic. If the linear source contains a clipped core, a different stretch cannot reconstruct it. If only the trial clips, return to the linear or previous-pass checkpoint and reduce or reshape the transformation.

Do not depend on an inverse stretch to replace saved originals. Clipping and some color-mode operations limit reversibility. Keep high-precision XISF checkpoints throughout.

Return to Target_MAS_trial made from the same linear-ready image and follow the MAS lesson. Compare the resulting display with STF off and at similar overall brightness. Do not run MAS on the already stretched GHS result for this comparison.

Inspect the same faint structure, smooth background, bright core, and star profiles at identical zoom. Include presentation size as well as 1:1. Avoid changing saturation, denoising, or HDRMT on only one branch during the initial comparison.

Prefer MAS when its adaptive baseline gets you to a satisfying balance with fewer adjustments. Prefer GHS when deliberate curve placement produces a balance you could not obtain as easily with the MAS trial. These are workflow preferences to test, not rules assigning a tool to every galaxy or nebula.

If both are poor, revisit the input’s gradients, noise, calibration, and actual dynamic range. More aggressive stretching can make defects easier to see without making the image better.

The preview looks washed out or too bright: check STF, active view, transformation type, and whether you accidentally reapplied a previous pass. Reset and compare from a known checkpoint.

SP changes seem extremely sensitive: use fine controls and inspect the source histogram and readout. A fixed numerical SP copied from another image is not a useful target.

Dark dust disappears: inspect shadows and blackpoint choices. Keep neutral endpoints for the baseline rather than clipping weak features to make the sky black.

Stars become harsh or lose color: lower strength, review mode and highlight handling, and compare the preserved stars branch if working starless. A stretched stars-only image needs its own restrained treatment rather than the nebula’s settings.

A predicted histogram disagrees with expectations: check the current view, masks, and color-mode options. The developer documentation notes limits to predicted-histogram behavior; inspect the applied trial itself before accepting it.

Save the accepted stretched image, linear source, each applied GHS instance in order, and comparison notes. Continue to masks, optional HDRMT or LHE, selective color work, and star restoration as appropriate. This is an original practice sequence; the process documentation governs current controls, while the older script primer supplies conceptual background.