Pedestal Removal

Remove a known residual constant electronic offset without confusing legitimate sky signal, noise, interpolation, or an isolated dark pixel with that offset. This page covers residual post-calibration offset handling only; bias, dark, flat, CFA state, and calibration-time output pedestals are upstream concerns covered by Calibration Foundations.

schema 1 · revision 2

Recognize it

Calibration records identify a residual constant offset that was not removed.
Calibrated subframes show left-edge clipping that requires a documented output pedestal during recalibration, rather than post-stack minimum subtraction.

Intended output

A still-linear image with only the confirmed uniform offset removed, no new shadow clipping, and recorded before/after channel statistics.

Required input state

Linear data with known calibration provenance.
A documented offset value or calibration diagnosis independent of the stack's single lowest pixel.
Before-operation channel minima, medians, and clipped-pixel counts recorded.

What NOVA does

NOVA skips remove_pedestal by default because a global minimum is insufficient evidence. The operation is declared with no tunable parameters and no Experiment Mode variants, and it requires capture or calibration provenance before it may be forced. If an operator explicitly forces the current function, it subtracts one global minimum uniformly from every channel and clips only as a numerical safety guard. The recorded M66 execution evidence below is from NOVA 1.24.7 and is evidence for that run and configuration, not a statement of current runtime identity.

Use when

Capture or calibration provenance establishes a remaining constant offset.
A controlled calibration diagnosis supplies the amount and the subtraction can be repeated on the original linear data.

Skip when

The only evidence is a positive minimum, median, or visually raised sky floor.
The image is an ordinary calibrated stack with no documented residual offset.
The proposed amount is a sky statistic rather than a calibration offset.
The data are nonlinear or the original calibration products are unavailable.

Limits

A positive global minimum does not identify an electronic pedestal.
Post-stack subtraction cannot restore values clipped during calibration.
Subtracting separate channel minima can alter color balance.
An output pedestal added during calibration to prevent clipping is not evidence that an arbitrary post-stack floor should be removed.
This step is not calibration. Bias, dark, and flat correction, CFA and debayer ordering, and calibration-time output pedestals all happen upstream of the stack this step receives; subtracting a scalar here is not a repair for any of them.

Scientific and aesthetic notes

Sky background is real signal plus noise; forcing its darkest sample to zero is not a calibration measurement.
Judge this analytically with statistics and provenance, not a before/after beauty slider.
If calibration clipped negative values, reprocess the subframes with an appropriate calibration-time output pedestal; later addition or subtraction cannot recover clipped structure.
Before-and-after statistics are diagnostics of what the subtraction did, not proof that it was warranted. The warrant comes from calibration provenance or a controlled calibration diagnosis, never from the statistics of the finished stack alone.
A future comparison of pedestal handling must start from the same calibrated source and vary only the confirmed offset treatment. Final-image aesthetic scores are not appropriate primary evidence for a calibration-domain decision.
The evidence status of this step is scientifically grounded, source-confirmed conditional behavior with historical execution evidence. It is not a general claim that ordinary SeeStar stacks contain a removable pedestal.

Evidence and measurements

Measurements	Acceptance criteria
Calibration provenance and the exact asserted offset value. Per-channel minimum, median, MAD/sigma, and clipped-pixel fraction before and after. Difference image confirming a spatially uniform subtraction.	The step remains skipped when no independent pedestal evidence exists. When applied, the difference image is spatially constant at the documented amount. No material increase in low-clipped pixels occurs. Channel-distribution shape and intended color relationships remain intact.

Cross-tool matrix

Tool	Equivalence	Expected result	Evidence
nova	Exact replay	Only the documented uniform offset moves; channel distributions retain their shape.	nova-pedestal-source, nova-m66-run-1.24.7
pixinsight	Algorithmically equivalent	A uniform shift by the confirmed amount without distribution-shape or color damage.	m66-manual-verification
siril	Algorithmically equivalent	Means and medians shift by the scalar while sigma/MAD and distribution shape remain stable.	siril-calibration-1.4.4, m66-manual-verification
saspro	Conceptual substitute	Each channel minimum moves to zero while spatial structure is translated by that channel's own amount.	saspro-source-1.18.0, saspro-synthetic-2026-08-14

NOVA Python pipeline current source-confirmed contract; execution evidence recorded at 1.24.7

Inspect capture and calibration provenance for an explicitly documented remaining constant offset.
Leave the step skipped when that evidence is absent; this is the default.
If force_apply is authorized from real provenance, record the input, uniform amount, and before/after channel statistics.
Read Calibration Foundations first when the suspected problem is clipping, vignetting, dust, or hot pixels: those are upstream calibration failures, and no post-stack scalar addresses them.

Controls and starting ranges

default action: skip
force_apply: false; true only with documented calibration evidence
subtraction amount: the confirmed offset; never infer from global minimum alone

Mask behavior: No mask: a true constant offset is a full-image calibration property.

Failure modes

A single cold pixel determines the amount.
Legitimate sky background is forced to zero.
Shadow clipping increases or channel balance changes.

Recovery

Revert to the unchanged linear input.
Recalibrate from subframes when the failure originated during calibration.

PixInsight PixelMath exact version not recorded; procedure tool-tested

Confirm a residual uniform offset from calibration provenance; otherwise skip.
In PixelMath subtract that measured scalar from $T with rescaling disabled.
Apply consistently to the intended channels and inspect Statistics before and after for minima, medians, and clipping.

Controls and starting ranges

expression: $T - confirmed_offset
rescale result: disabled
offset: measured calibration value; no generic starting number

Mask behavior: No mask for a confirmed full-frame constant offset.

Failure modes

Using min($T) as proof instead of merely as arithmetic.
Using a sky median as the offset.
Per-channel subtraction changes color balance.

Recovery

Undo and return to the original linear image; recalibrate if clipping occurred earlier.

Siril Pixel Math and calibration 1.4.4 documentation; procedure tool-tested in 1.4.2

Prefer correct bias/dark/flat calibration and inspect the calibration statistics before considering a later subtraction.
Skip post-stack subtraction unless provenance confirms a residual offset.
For the demonstrated manual arithmetic, subtract one confirmed scalar in Pixel Math and compare channel statistics before and after.

Controls and starting ranges

post-stack scalar: confirmed offset only; no generic starting number
display normalization: disabled while comparing numeric statistics

Mask behavior: No mask for calibration-wide offset handling.

Failure modes

Treating the lowest channel minimum as an identified pedestal.
Clamping a large part of the real sky background to zero.

Recovery

Undo; return to the calibrated linear input and diagnose the calibration chain.

Seti Astro Suite Pro Pedestal Removal 1.18.0 source-inspected and synthetic-array tested

Treat target class as irrelevant: decide from calibration provenance and linear-image statistics.
Skip unless each channel's residual offset is independently established; a positive minimum alone is not evidence.
If applied, record each channel minimum and compare distributions, clipping, and color before and after.

Controls and starting ranges

operation: per-channel minimum subtraction; no adjustable amount
default decision: skip without independent per-channel calibration evidence

Mask behavior: The active full image or preview is processed; this is not a calibration mask operation.

Failure modes

A cold pixel or legitimate sky sample defines a channel's subtraction.
Different channel minima alter the established color balance.
Clipping or earlier calibration damage is mistaken for a removable pedestal.

Recovery

Undo immediately and return to the unchanged linear image or recalibrate the source frames.

Validation status

Not yet Jeff-validated. Use this as sourced guidance, not a certification.

Sources

M66 manual recipe verification feedback M66_Manual_Recipe_Verification_Feedback.md — Jeff validated
NOVA pedestal implementation and conditional gate nas_server/seti_astro.py:3096; nas_server/auto_process.py:655 — NOVA source-confirmed
M66 workflow 1.24.7 execution record critiques/20260802_045251_seestar_galaxy.json — NOVA execution record
Siril 1.4.4 calibration documentation — Vendor-documented
Installed Seti Astro Suite Pro 1.18.0 processing source saspro/pedestal.py; saspro/numba_utils.py; saspro/stacking_suite.py; saspro/abe.py; saspro/sssc.py — Artifact-confirmed
SASpro 1.18.0 synthetic-array operation checks 2026-08-14 local validation: pedestal, cosmetic correction, and ADBE — Tool-tested
Handbook concept page: Calibration Foundations handbook:calibration-foundations — NOVA source-confirmed

Related evidence

The M66 recipe and Jeff-validation ledger will link here when their static exports are available. Until then, this article remains explicitly unvalidated.