Separate Stars with StarXTerminator and Restore Them with ImageBlend

Stars and extended objects often need different processing. Separating them lets you refine a nebula or galaxy without making every star brighter, and adjust the star field without applying those changes to the background object.

StarXTerminator, from RC Astro, produces a starless image and can generate a separate stars image. ImageBlend, from Cosmic Photons, provides blending controls for combining images. StarXTerminator requires its own purchased or trial license; ImageBlend is available from its publisher without that license requirement.

This lesson uses a broadband RGB image and linear star separation, then restores stretched stars over a stretched starless image. Keep an untouched original: separation is an estimate, and the stars output may contain features you wanted to keep in the nebula or galaxy.

Use a calibrated, integrated RGB image. Complete operations that need the full star field, such as plate solving, SPFC/MGC when used, and broadband SPCC, before removing stars. Keep this checkpoint linear, with no permanent stretch. STF can be active for display.

Save Target_RGB_before_stars.xisf and make a working duplicate. Check that no unintended mask is attached to the working image. For a first exercise, avoid mixing stars from another target, camera, crop, or narrowband branch.

Install the two tools using the setup guides under Continue learning. Open StarXTerminator from Process → RCAstro and ImageBlend from Script → Utilities, or their equivalent menu locations in your version.

For linear input, RC Astro recommends subtraction-based extraction: leave Unscreen Stars off. Unscreen is intended for already stretched input. Do not turn it on merely because the later blend mode is Screen.

Blink or compare the original and starless image at matched zoom and display settings. Inspect bright-star locations, diffraction spikes, faint stars, compact nebular knots, and small galaxies. Look for leftover stars, holes, streaks, or damaged structure.

Inspect the stars image too. It should predominantly contain stars and their profiles, rather than obvious filaments or galaxy detail. If important structure appears there, pause before strengthening or reducing that layer: the separation needs review.

RC Astro advises preserving the stars view’s inherited STF rather than running a new STF Auto Stretch on it. An independent auto-stretch can greatly exaggerate tiny residual background values. Record or save that display setup if you intend to use it for a permanent star stretch.

If the result is poor, return to a fresh source copy. Consult the current usage notes for star-profile and processing-state limitations; a different extraction stage may help, but it should be tested rather than assumed better. Never erase a compact feature just because it resembles a star.

Keep the linear pair as checkpoints. Work on duplicates named Target_starless_stretched and Target_stars_stretched. Complete any intended linear-only work before applying permanent stretches.

For the starless branch, use the MAS lesson to make a restrained permanent stretch and inspect faint structure. For the stars, start with a gentle midtones stretch. If using the inherited STF as a starting point, transfer its instance to HistogramTransformation, inspect the resulting histogram settings, then apply HistogramTransformation once to a stars duplicate. Reset STF afterward to judge the actual result.

The goal for the stars branch is readable stars with preserved color and profiles, not a bright background. Adjust its permanent stretch from a fresh linear duplicate when necessary. Avoid forcing a background target designed for a nebula onto a mostly black stars image.

Keep the branches registered: identical dimensions, crop, orientation, and pixel alignment. Apply any geometry change consistently to both. Matching width and height alone does not guarantee alignment.

Screen blending brightens the base where the stars contribute. In normalized values, it follows 1 − (1 − base) × (1 − stars). It is different from ordinary addition. After separate stretches and edits, this creates a finished composite; it does not promise an exact reconstruction of the original linear image.

Start with the full-opacity result as a baseline. If the stars dominate, try a small opacity reduction, such as 0.90 when the control uses a 0–1 scale. This is a comparison value, not a recommended final strength for every image.

Reducing opacity lowers the stars’ contribution; it is not the same as shrinking their profiles. For bloated stars, review the stars stretch and extraction quality before reducing opacity until they become unnaturally faint.

Change one control at a time and generate separate trial images. Keep the base fixed while comparing star strength. Avoid adding filter, color, and midtone changes simultaneously, because it becomes difficult to identify which change helped.

Inspect faint stars as well as bright ones. The preferred result should support the object while still looking like a coherent star field. Record the retained settings rather than treating a lower star count as the goal.

With STF off, examine bright stars, crowded areas, faint corners, and star positions over bright nebulosity at 1:1. Look for doubled stars, dark rings, colored halos, holes, hard white cores, or newly visible remnants.

Compare source, starless, stars, and composite before accepting a fix. A halo can originate in the optics, separation, stretching, or the blend; changing opacity may only conceal it. If small galaxies or nebular knots disappeared earlier, bringing back a dimmed stars layer may not restore their intended appearance.

Save the accepted composite as Target_with_stars.xisf, along with both branches and the pre-separation image. Record the extraction state, Unscreen choice, model version, branch stretches, ImageBlend mode, opacity, and other changes.

Keep the XISF checkpoints for future work, then export a separate display copy when needed. The composite is a nonlinear finishing checkpoint; return to the preserved linear source for another calibration or extraction experiment.

StarXTerminator also accepts nonlinear data. Review the current vendor instructions and consider Unscreen Stars for extraction intended for Screen restoration. Preserve an unedited copy of that stretched source and record the extraction method.

Do not permanently stretch the resulting stars again just because the linear workflow above includes a stretch step. First inspect the outputs with STF off and establish which operations have already changed their pixels.

A Screen/Unscreen pair can support restoration of a stretched source, but separate edits and removal artifacts can still change the outcome. Compare a baseline recombination before making creative adjustments. Do not assume that different extraction methods or branch states are interchangeable.

No stars output: verify Generate Star Image and rerun on a fresh source duplicate. Do not rerun extraction on the already starless result to recover the original star field.

Double stars or shifted halos: check registration, crop, rotation, and image dimensions. A blending mode cannot correct misalignment.

Washed-out composite: reset STF, check that both inputs have the intended permanent stretch, confirm Screen mode, and inspect whether a second background was introduced through an over-stretched stars layer.

Grid or tile artifacts, missing stars, or removed subject detail: use the current StarXTerminator documentation to evaluate options such as Large Overlap where applicable, and compare fresh runs. Do not adopt every option as a default fix.

This is an original practice sequence using vendor usage notes and the publishers’ descriptions. ImageBlend’s built-in documentation is the reference for controls in the installed version. The linked Adam Block overview identifies Screen blending as a star-management technique; its full training content requires purchase and is not reproduced here.