Description
A properly preserved star shows a dense core at peak brightness surrounded by a radial gradient that reveals its temperature: blue for hot stars (type O, B), white-yellow for solar-type stars (G), orange-red for cool stars (K, M).
When the core is "clipped", it becomes a pure uniform white disk: color information is gone and internal dynamic range is lost. The defect reflects highlight clipping (saturation at 1.0), most often introduced by an overly aggressive stretch on the highlights, and sometimes by direct saturation at acquisition (individual exposure too long on bright stars).
Once information is clipped, it is permanently lost: no tool can reconstruct the internal gradient of a star whose central pixels all read 65535.
Preserving stellar colors is one of the visible hallmarks of a carefully processed image; conversely, clipped stars are an immediate signature of a heavy-handed workflow.
Visual signature
Bright stars display a perfectly white core, with no nuance or gradient, often extending 5 to 20 pixels in diameter depending on the severity of the clipping.
The transition between the white core and the colored halo is sharp rather than gradual. Blue and red stars lose their chromatic identity at the center, becoming visually identical to white stars.
On the RGB histogram, a flat peak at 1.0 (or 65535 in 16-bit) appears on all three channels.
On close inspection, the saturated core spills into saturated spikes on very bright stars (Sirius, Vega, Aldebaran in wide-field), accentuating the "white blob" effect with no internal structure.
On an overly stretched image, the effect extends even to moderately bright stars.
Differential diagnosis
Do not confuse with acquisition saturation (star already clipped in the raw frame, before any processing), which is diagnosed by checking the histogram of a single light: if already saturated at the source, the defect originates in the exposure, not the stretch.
Distinguish from bleeding/blooming on older CCD sensors (saturation bleed flowing vertically along a column, virtually absent on modern CMOS).
Different from over-deconvolution, which produces shrunken stars with a dark ring, not a clipped core.
Do not mix up with a optical chromatic defect (colors shifted but present, whereas here they are absent).
Also check whether the entire image shows highlight clipping (other saturated white areas: galactic nuclei, the core of M42) or whether clipping is limited to stars: the diagnosis and solution differ depending on the extent.
Distinguish from over-saturation: here the star core is clipped in luminance (pure white); there it is color that is pushed to excess. A clipped star loses its hue through luminance clipping, not through excessive color saturation.
Probable causes
- A single overly aggressive stretch (HistogramTransformation pushed to saturation)
- ArcsinhStretch with stretch factor set too high without monitoring highlight clipping
- GHS (Generalized Hyperbolic Stretch) with a curve poorly calibrated on the highlights
- No star protection before stretching (no prior StarXTerminator/StarNet run)
- Stacking images already saturated at acquisition (bright stars already clipped in the lights)
- Individual exposure too long on targets with very bright stars (Pleiades, Orion Trapezium without HDR)
- RGB combination without normalization, one channel pushing stars to saturation
- ColorSaturation applied after clipping has already occurred (reinforces the loss)
- Conversion to 16-bit then JPEG, which finalizes and locks in the clipping
Course of action
- Separate stars and background with StarXTerminator before any aggressive stretch, then process each independently
- On the separated stars, apply a very gentle stretch (light ArcsinhStretch or conservative GHS)
- Monitor RGB statistics during the stretch: peaks must stay below 0.95 in the highlights
- For targets with bright stars (M42, Pleiades, M45), plan HDR exposures: short (5-30s) combined with long (180-300s)
- In PixInsight, HDRComposition combines multiple exposure lengths to preserve the cores of bright stars
- Enable "preserve highlights" protection in HistogramTransformation when available
- For already-clipped stars, RepairedHSV or ColorCalibration applied at boundary pixels may recover some hue
- Work in 32-bit floating point throughout the workflow to avoid premature clipping
- After StarXTerminator processing, recombine unclipped stars with the aggressively stretched background
The Doc's advice
A pure white star simply does not exist in nature: even Sirius (type A, blue-white) has a tint. If your stars all come out white as pearls against a black background, you have clipped them. The reflex to build in: run StarXTerminator before any serious stretch. You protect the stellar dynamic range, and you can then pull the background as hard as you like without touching star colors. It is probably the single biggest quality-to-time gain in the entire modern workflow.
Think you can see this defect in your image?
Run a diagnosisFrequently asked questions
Can the color of already-clipped stars be recovered?
Very partially. If clipping is complete (all three RGB channels at 1.0 in the core), the information is gone and no algorithm can reconstruct it. If only one or two channels are saturated, tools such as RepairedHSV (PixInsight) can approximately reconstruct the hue from the unclipped channels. For a satisfying result on an existing image, the most pragmatic path is often to replace clipped stars with synthetic stars generated via StarNet++ or a library of calibrated photometric stars.
Is HDR imaging always necessary to preserve stars?
No, not for most targets. On galaxies or faint nebulae in fields with moderate stars, uniform exposures processed with StarXTerminator and a separate stretch are more than sufficient. HDR exposures become necessary on targets with very high dynamic range: the Orion Trapezium, the core of M31, the Pleiades in wide-field, IC 410 (Tadpoles). The practical rule: if your individual exposures saturate the bright stars in the field, add a set of short frames (5-30s, 20-30 frames) to merge in.
Does StarXTerminator definitively solve the problem?
It largely eliminates it, yes. By separating stars from the background before the main stretch, one can apply a hard stretch to the nebulosity without touching the stellar dynamic range. The stars, processed separately with a gentle stretch, retain their colors intact. This has become the default workflow in PixInsight and Siril since 2022. One important limit: if the stars are already clipped in the pre-StarX master (because they saturated at acquisition), no tool will save them.
Is highlight clipping on a few pixels really a problem?
On truly very bright stars (magnitude <3 in the field), a saturated core over 1-3 pixels is unavoidable and visually acceptable; the eye interprets that ultra-bright point as normal. The defect becomes critical only when clipping extends to many stars or to cores larger than 5-10 pixels, which visibly erases chromatic information. A good reference: in a well-processed image, the majority of magnitude 6-10 stars in the field should be distinguishable by color, even if magnitude 1-3 stars have a punctually saturated core.