Exposure Correction

Purpose and expected result

Exposure Correction multiplies normalized samples or one selected RGB-derived component by a scalar and clamps the result to [0,1]. In linear light, a positive scalar can model a brighter exposure and a scalar below one can model a darker exposure. The Exposure control is not an adjustment in stops: it does not use the standard stop equation 2^e, and its tooltip states the factor it applies.

For entered value e, AfterNight computes the float multiplier m as

\[ m(e)= \begin{cases} 1.02 + e/5, & e<0,\\ 1, & e=0,\\ 1 + e/5, & e>0, \end{cases} \qquad y=\operatorname{clamp}(m(e)x,0,1). \]

Thus +1 means 1.2x, not the 2x of a true +1-stop adjustment; only the positive endpoint +5 happens to produce 2x. Negative values use an extra 0.02 offset, so the mapping is discontinuous at zero and small negative values can brighten rather than darken. The guide reports this current behavior so that results are reproducible; it is not a recommended stop definition.

For unclipped RGB in RGB mode, the common multiplier preserves stored channel ratios. Once one channel clips, color and highlight ratios change. Selecting one stored or converted component changes only that component before RGB reconstruction and is not a physical exposure adjustment.

Every Apply records its result as nonlinear on the processing-state cards, even an unclipped scaling of linear data, and keeps the source's other FITS cards, astrometric solution, and acquisition record; see Inputs, outputs, and metadata.

At a glance

Item	Current behavior
Menu path	Process > Adjustments > Exposure
Enabled when	An image or sequence is active, the same process is not already attached to it, and no attached process holds exclusive access
Inputs	One intended normalized grayscale or RGB still or sequence; an optional active mask must match target geometry
Outputs	Destructively replaces a still or, after successful temporary processing, a sequence; Take snapshot creates an auxiliary still without replacing the target; component modes reconstruct RGB; automatic sequence CFA debayer can change frame format/channels; no file is written directly
Recommended image state	Linear-light data for exposure-like scaling; on nonlinear data this is only encoded-value brightness scaling; either way the result is recorded as nonlinear
Preview support	Asynchronous RT preview of the full image or a selected preview area, using the same multiply-and-clamp operation as Apply, plus a histogram of the mask-aware result; the preview viewer can display Linear values or a display-only Autostretch
Sequence support	Supported: each decoded frame is processed exactly as a still image of it would be, into a temporary sequence that replaces the target only after every frame succeeds, with per-frame progress and Cancel; the configured CFA auto-debayer request can produce RGB frames
Mask support	Supported for preview, still Apply, and sequence Apply by post-transform blending; white applies, black protects, gray blends, and Invert Mask reverses weights
External dependencies	None; bundled engine only; no tool, profile, catalog, or network service

Before you start

• Use linear-light data if the goal is exposure-like scaling. Multiplying gamma-encoded or already stretched samples changes display values, not scene exposure.
• Apply records its result as nonlinear even when no sample clipped. If later steps rely on a recorded linear state, run them before Exposure Correction.
• Treat the control as a project-specific scalar control, not a stop control. For a true multiplier 2^s, the current positive branch can represent only 1..2x, and its label does not identify the equivalent stop value.
• Inspect highlights before Apply. Every stored/component value above 1/m is clipped to 1; no later operation can recover those distinctions.
• RGB preserves ratios only while all channels remain below the clamp. Red/Green/Blue or derived-component processing intentionally changes color.
• Avoid hue/chroma modes for ordinary exposure work. Their normalized encoded coordinates are multiplied and clamped as ordinary scalars; hue is not treated as a circular angle.
• In a converted-component mode only the component is clamped: on a floating image the reconstructed RGB color is not clamped to [0,1], and on an integer image it saturates at the type's range.
• For a sequence, verify the selected Channel, active mask, and CFA auto-debayer setting before Apply. The same captured Exposure/Channel selection is used for every decoded frame; cancellation discards the temporary output.

Procedure

1. Select a still or sequence and open Process > Adjustments > Exposure. Confirm Target Image, leave Channel: RGB, and keep Preview selected.
2. Change Exposure slowly. Positive values multiply by 1+e/5; negative values multiply by 1.02+e/5. The identity value 0 disables Apply.
3. Inspect the histogram's white endpoint and bright colored stars at 100%. Reduce the value if one or more channels form a clipped plateau. On a large image, check Preview area on the preview view to tune on one region.
4. If choosing a single stored/derived channel, inspect color and gamut. This is component scaling, not exposure.
5. If a mask is active, judge the mask-aware preview. Histogram clipping refers to displayed result bins after blending.
6. Apply when the result is acceptable. For a sequence, monitor the progress dialog and use Cancel to preserve the original sequence. Image History holds one Exposure Correction step that Undo reverses, and the result is recorded as nonlinear with exposure-correction as the operation.

Complete parameter and action reference

UI label/action	Type and unit	Runtime default	Persisted-default exception	Valid values, step, or constraints	Effect of increasing, decreasing, or selecting	Dependencies and enable/visibility rules	Quality, artifact, runtime, memory, or disk trade-off	Scope: preview/apply, channel, mask, still/sequence	Persistence
Histogram plot	Graph, normalized intensity/count	Source or current preview; RGB channels overlaid	None	Binning comes from Plot resolution	Shows source or mask-aware preview distribution and approximate endpoint-bin clipping; never changes samples	Show histogram controls visibility	Recalculation adds work; endpoint counts are resolution-dependent	Display only; shown source/preview channels	Recomputed after source, preview, or resolution changes
Left-drag in histogram	Graph gesture	No readout	None	Hold left button and move	Shows normalized coordinate readout	Histogram must be shown	Diagnostic only	Display only	Ends on release
Middle-drag in histogram	Graph gesture	Fitted/unpanned	None	Hold middle button and drag	Pans within graph scroll area	Useful only beyond viewport	No output effect	Display only	Until reset/window close
Ctrl+mouse wheel in histogram	Graph gesture	Fitted size	None	Up 1.20x; down 0.8x; plain wheel scrolls	Zooms around cursor	Interactive graph enabled	Rendering area only	Display only	Window lifetime; process Reset does not change it
Zoom in	Toolbar action	Fitted size	None	Each action 1.30x	Enlarges plot	Disabled with hidden histogram	Rendering only	Display only	Window lifetime
Zoom out	Toolbar action	Fitted size	None	Each action 0.70x	Shrinks plot	Disabled with hidden histogram	Rendering only	Display only	Window lifetime
Zoom reset	Toolbar action	—	None	One action	Fits plot/resets scrolling	Disabled with hidden histogram	Rendering only	Display only	Immediate
Show grid	Toggle	Checked	None	Checked/unchecked	Shows/hides grid	Disabled with hidden histogram	Display-only aid	Display only	Window lifetime; hiding histogram resets grid internally
Show clipping info	Toggle	Checked	None	Checked/unchecked	Shows/hides histogram endpoint-bin counts	Disabled with hidden histogram	Binning-dependent estimate, not a count of values clipped specifically by this process	Display only	Window lifetime
Absolute scales	Toggle	Unchecked	None	Checked/unchecked	Checked normalizes each visible histogram channel to its own peak	Disabled with hidden histogram	Can conceal relative channel strength; no image normalization	Display only	Window lifetime
Draw histogram areas	Toggle	Unchecked	None	Checked/unchecked	Filled-area versus line drawing	Disabled with hidden histogram	Area overlap can obscure channels	Display only	Window lifetime
Show histogram	Toggle	Checked	None	Checked/unchecked	Hides/clears graph and disables toolbar, or reloads source/preview	Always available while idle	Hiding saves graph work, not preview processing	Display only	Window lifetime; process Reset does not change it
Plot resolution	Combo box, bins	8-bit (256)	None	6-bit (64), 7-bit (128), 8-bit (256), 9-bit (512), 10-bit (1024), 12-bit (4096), 14-bit (16K), 16-bit (64K), 100 levels, 1000 levels, 10000 levels	More bins refine the graph/endpoint estimates	Disabled with hidden histogram	Higher bins cost graph memory/rendering; output precision unchanged	Display only	Window lifetime
Exposure	Double value plus slider labelled Exposure; its tooltip states the factor below	0.00 identity	None	-5.00..+5.00; two decimals; spin step 0.03; slider linear over 1,000 positions	Positive uses 1+e/5; negative uses 1.02+e/5; -0.10 is another identity and (-0.10,0) brightens despite being negative	Enabled while idle; Apply disabled only exactly at 0.0	Positive values clip samples above 1/m; negative branch reaches 0.02x at -5; displayed value is not a true stop count	Preview/Apply; selected channel/component; mask post-blend; still/sequence	Captured per execution; Reset/success returns 0.0
Target Image	Image selector	Attached/active target	None	Open eligible image or sequence	Changes source, Channel list, histogram, Preview, Apply	Active image required; duplicate/exclusive gates apply	Wrong target is destructively replaced; sequence writes a temporary result and replaces only after success	Preview/Apply source and active mask	Target lifecycle; Reset does not change it
Channel	Combo box	RGB for RGB; disabled Grayscale for mono	None	RGB, Red, Green, Blue, CIE L* (Lightness), CIE a, CIE b, CIE c, CIE h, CIE X, CIE Y (Luminance), CIE Z, HSL Hue, HSL Saturation, HSL Lightness, HSV Hue, HSV Saturation, HSV Value, YCbCr Luma, YCbCr Blue Chroma, YCbCr Red Chroma	RGB/Grayscale multiplies stored samples; Red/Green/Blue multiplies one stored channel; other modes convert, multiply one encoded component, and reconstruct RGB	Component choices require RGB; mono selector disabled	RGB preserves ratios until clipping; other modes add conversion work and gamut risk (a floating reconstructed color is not clamped to [0,1], an integer one saturates at the type's range); source metadata is kept in every mode	Preview/Apply; selected component; mask post-blend; still/sequence	Captured by current selection; Reset/success returns RGB when target is RGB
Preview	Checkbox	Checked	None	Checked/unchecked	Enables/disables the asynchronous multiply/clamp preview and switches histogram result/source	Process idle	Result/histogram buffers cost time/memory; same mathematical transfer as Apply	Preview only; mask-aware; current still/current displayed sequence frame	Window lifetime; Reset does not change it
Preview pane	Full-image viewer	Current result	None	Standard viewer zoom/pan/display controls	Shows mask-aware result	Exists only with Preview	Holds result image; component conversions add buffers	Preview only; target unchanged	Window lifetime
Display Mode	Preview-toolbar combo box	Inherits the target view's Linear or Autostretch mode when the preview is created	None	Linear, Autostretch	Changes only how the Exposure result is displayed; Autostretch uses cached full-target source statistics without changing process output pixels	Exists only with Preview; Autostretch enables Autostretch Settings	A source-derived display stretch can reveal faint structure or exaggerate noise and must not be mistaken for Exposure output	Preview display only; no effect on Apply, snapshots, masks, or sequences	Preview-view state; process Reset does not change it
Autostretch Settings	Preview-toolbar dialog action	Disabled in Linear mode; current target-view settings are inherited	Current shared-view values load when opened	Opens the shared settings for Target Background, Shadows Clipping, linked channels, neutralization/color calibration, saturation/background, brightness, contrast, banner, Reset, and Close	Tunes only the preview viewer's display stretch	Enabled only in Autostretch mode; the action is disabled while its dialog is open; Apply Stretch is hidden in this RT-preview dialog	Aggressive display settings can exaggerate noise/color without changing Exposure output	Preview display only	Viewer-scoped; not captured in Exposure parameters or history
Switch image	Preview-toolbar action	Exposure preview shown	None	Click; Space; right-click; Ctrl+Shift+O	Toggles the preview viewer between the current result and original target	Exists only with Preview	Display comparison only; no recomputation or pixel change	Preview display only; still/current displayed sequence frame	Preview-window lifetime
Preview area	Preview-toolbar checkbox	Unchecked	None	Checked/unchecked; move/resize the selection overlay	Limits preview and snapshot processing to a selected area and composites that result over the original full image; inside the area the preview is what Apply makes of an image of that area	Exists only with Preview	Reduces multiply/component work while tuning; does not limit final Apply	Preview and Take snapshot only; mask is cropped consistently	Preview-window lifetime; Reset does not change it
Take snapshot	Preview-toolbar action	No snapshot	None	Click or Ctrl+Shift+K	Re-executes the current Exposure settings at source depth and opens a new still named <target> - Exposure Correction - SNAPSHOT	Exists only with Preview	Allocates and processes an auxiliary image; active mask and selected preview area are honored; Display Mode/Autostretch is not baked into the snapshot	New still image; target is unchanged; a snapshot is a preview result and is not recorded as nonlinear	Each action creates a separate image view
Apply	Button/action	Disabled at Exposure 0.0	None	One action at any other displayed value, including the second identity -0.10	Captures Exposure and runs one fused multiply-and-clamp program over the selected channel for a still or each decoded sequence frame	Target must lock; controls disabled during the run	One parallel engine pass; integer samples add a floating working copy; conversions and masks add passes; recording the nonlinear state usually copies the result once more; clipping is irreversible	Destructive still/sequence replacement; active mask applies	Success records one history step, then resets Exposure and Channel
Reset	Button/action	Enabled while idle	None	One action	Restores Exposure 0.0 and RGB for an RGB target	Does not change Target, Preview, or histogram toolbar/zoom	No output cost	UI only	Immediate
Close	Button/action	Enabled while idle	None	One action	Closes without Apply	Disabled during commit	Pending preview result is discarded	UI only	No setting persistence
Sequence progress / Cancel	Shared modal progress dialog and Cancel action	Created automatically when sequence Apply starts	None	One progress update per frame; cooperative Cancel	Processes decoded frames into a temporary sequence and replaces the target only after success	Sequence target only	Sequence processing adds decode/encode and temporary-disk cost; automatic CFA debayer can produce RGB frames	Sequence Apply; chosen channel and active mask apply to every frame	Cancel or a failed frame discards the temporary sequence; success replaces the sequence, records one history step, and resets settings

How the algorithm works

Transfer, clipping, and channel behavior

The standard exposure-stop law is m=2^s: +1 doubles linear-light samples and -1 halves them. AfterNight instead uses the piecewise linear control map shown above. On the positive branch, the equivalent stop change is log2(1+e/5); on the negative branch it is log2(1.02+e/5).

For normalized finite input, multiplication is followed by an unconditional clamp. Positive values therefore create a hard highlight plateau wherever x >= 1/m. The negative branch's multiplier remains positive over its allowed range, reaching 0.02 at -5, so ordinary normalized values do not black-clip except exact zero. Floating samples outside [0,1] are also clamped after multiplication; a not-a-number sample stays not-a-number. Samples that the image's own validity rule or mask marks as invalid are left as they were.

RGB and selected components

• RGB uses one multiplier independently on every stored R/G/B value. Ratios are preserved before clipping; they diverge when channels reach white at different input levels.
• Red, Green, or Blue changes only one stored channel.
• CIE, HSL, HSV, and YCbCr modes convert the image's red, green, and blue, multiply and clamp one component handed over in the unit interval as a Float32 working plane (Float64 for a 64-bit float image), then reconstruct RGB. A source whose description says it is linear is read as linear RGB; any other is read as sRGB-encoded. Reconstruction can leave the gamut or alter hue/chroma: a floating result is not clamped to [0,1], and an integer result is rounded to the nearest code and saturated at the type's range. CIE components are reconstructed through the exact inverse of the XYZ matrix, and YCbCr through unrounded luma-chroma coefficients. Hue receives no circular wrap treatment.
• Grayscale and one-channel CFA stills are treated as one stored channel. This process does not demosaic a still.

The operation keeps the source pixel type. Integer input is read in normalized units (an unsigned integer as a fraction of its largest value, a signed one counted from its smallest value to its largest), converted to a floating working copy (Float64 for the two 32-bit types, Float32 otherwise), multiplied by the single-precision factor, clamped, and rounded back to the nearest code, ties to even. Float input remains float and double remains double. The multiply and the clamp are one fused engine pass on the operation's threads. Preview and Apply use the same engine operation, although histogram bins remain diagnostic only.

Inputs, outputs, and metadata

Image state, metadata, masks, and history

Scalar multiplication is exposure-like only in linear light and, before clipping, preserves linearity. Every Apply nevertheless records its result as nonlinear on the processing-state cards a saved file carries: ANLIN becomes nonlinear, ANASRC process, ANOP exposure-correction, and ANBASIS the basis a stretched image of its kind is in (for example display-rgb for an RGB image). A linear transfer function the source claimed is no longer claimed, and the image's lineage gains an entry naming the engine operation with the exposure and its factor.

• Every other FITS card, the astrometric solution, and the acquisition record are kept, in stored-channel and converted-component modes alike.
• A Take snapshot image is a preview result and is not recorded as nonlinear.

An active mask is blended afterward as source*(1-mask) + corrected*mask, reversed for an inverted mask. Factor derivation and clipping happen before this blend. Image History records one step named Exposure Correction, described as Adjust exposure to <e>, followed by in channel: <Channel> for an RGB target (for example Adjust exposure to 1.5 in channel: RGB); it states the displayed control value, not the actual multiplier/equivalent stops. Undo restores the previous image. Successful still Apply resets Exposure to zero and Channel to RGB.

A run whose stop is requested (Cancel in a sequence's progress dialog, or the target closing while a run works) is stopped inside the engine operation and publishes nothing; the image and its history stay as they were.

Sequence Apply processes each decoded frame as an Apply processes a still image of it, active mask included, writes the frames to a temporary sequence, supports Cancel from its progress dialog, and replaces the target only after every frame succeeds; a cancelled or failed run leaves no temporary output. The current CFA auto-debayer request is passed to the sequence reader. One history step is recorded.

Parameter interactions and starting points

Combination	Conservative starting point	Too weak looks like	Too strong or defective looks like	Adjustment order
Exposure + highlight headroom	Small positive increments around 0.1..0.3	Brightness barely changes because multipliers are only 1.02..1.06	White plateau and colored-star saturation	Inspect maximum/channel histograms, raise slowly, stop before clipping
Negative value + current discontinuity	Use e <= -0.10 only after checking the formula	-0.10 changes nothing	Values between -0.10 and 0 brighten; near zero the preview jumps	Treat the value as scalar control, not stops; verify result numerically
RGB + channel clipping	Leave RGB selected	Desired brightening remains modest	One channel clips before others and changes color	Inspect each RGB endpoint, then reduce Exposure
Component mode + color	Prefer RGB for physical exposure intent	Selected component has little visible effect	Hue, saturation, or gamut shifts after reconstruction	Choose component only for deliberate color editing; inspect colors
Mask + clipping	Use soft transitions and modest positive values	Protected region changes too little	Applied region clips or mask boundary becomes visible	Tune unmasked transfer, then inspect mask-aware result

Performance

The core is one parallel multiply/clamp pass. Integer samples add a floating working copy; stored single-channel modes avoid color conversion; component modes compute a floating working plane and convert back. The run works on a copy of the target, and recording the nonlinear state usually copies the result once more; if the session's memory budget refuses one of these copies, Apply publishes nothing and reports the refusal. An active mask adds another full-image blend, and Preview also retains a result and updates the histogram. Histogram resolution changes diagnostic rendering, not output precision or the multiplier.

Disable Preview while entering several values on a large image, or use Preview area, then re-enable the full preview for the final check. No approved project image was timed, so no production throughput or memory figure is claimed.

Worked example

• Evidence mode: Source-, focused-test-, and calculator-derived; not executed on project-owned astronomical data.
• Positive setting: With Exposure +1, the current multiplier is 1.2, not 2. Inputs [0.25, 0.50, 0.90] become [0.30, 0.60, 1.00]; the last raw value is 1.08 and clips. The equivalent unclipped change is about log2(1.2)=+0.263 stops.
• Negative setting: With Exposure -1, the multiplier is 0.82, so input 0.50 becomes 0.41. This is about -0.286 stops, not -1 stop (0.25).
• Discontinuity checkpoint: Exposure -0.10 produces multiplier 1.0 even though Apply is enabled. Exposure -0.05 produces 1.01, so input 0.50 becomes 0.505: a negative displayed value brightens. At exactly 0, the multiplier returns to 1 and Apply is disabled.
• Inspect in real use: Check each RGB white endpoint, the brightest star cores, and the recorded history value. Do not translate the displayed value into stops without applying the piecewise equation.
• Limitations: No real-image color fidelity, clipping recovery, metadata round trip, executed sequence output, or production timing was measured.

Warnings and limitations

The current control is not calibrated in exposure stops

AfterNight does not compute 2^Exposure. A displayed +1 gives 1.2x, and -1 gives 0.82x. Use the documented piecewise multiplier when exact scaling matters.

Small negative values can brighten

The negative branch adds 0.02, creating a discontinuity at zero. -0.10 is an enabled identity and values between -0.10 and 0 brighten.

• Positive scaling hard-clips every value above 1/m; clipped detail and color ratios cannot be restored later.
• Scaling nonlinear/encoded samples is not scene exposure compensation. Work in linear light when that interpretation matters.
• Every Apply is recorded as nonlinear, even an unclipped scaling of linear data. Run steps that rely on a recorded linear state before Exposure Correction.
• Component modes are color-coordinate edits, not exposure changes, and a floating reconstructed color is not clamped to [0,1].
• History records the displayed value rather than actual multiplier, so the piecewise formula is required to reproduce the math outside AfterNight.
• Integer results are rounded to the nearest code, ties to even.
• A not-a-number floating sample is not repaired. Use finite normalized data.

Troubleshooting

Symptom or exact error	Verified cause/validation path	Recovery	Evidence source
+1 does not look one stop brighter	Current multiplier is 1+1/5=1.2, not 2^1=2	Use the current formula, which the tooltip states, and judge highlights; do not read the value as stops	ExposureCorrectionProcess::exposureFactor; worked calculator reproduction
A small negative value brightens	Negative branch is 1.02+e/5; it exceeds one for -0.10<e<0	Use e<-0.10 for darkening and verify the actual multiplier	ExposureCorrectionProcess::exposureFactor
Exposure -0.10 runs but changes no pixel in [0,1]	It is a second mathematical identity, while Apply is disabled only at exact 0.0; floating samples outside [0,1] are still clamped, and the result is still recorded as nonlinear with a history step	Reset to 0; avoid committing -0.10	ExposureCorrectionProcess::exposureFactor; ExposureCorrectionProcess::settingsUpdatedImpl
Highlights become flat white or change color	Multiply result is clamped to one; RGB channels can reach the clamp at different inputs	Reduce Exposure or create headroom before scaling	buildScaleBiasProgram clamp; ExposureCorrectionProcess::executeTypedProcessIteration
A floating image has values outside [0,1] after a component mode	Only the component is clamped; it is converted back to RGB without clamping a floating result	Use a smaller change or a stored-channel mode	applyProcessChannelScope; ChannelScope.h recomposition
A later step treats the result as nonlinear	Apply records ANLIN nonlinear and ANOP exposure-correction on the result, even without clipping	Expected; run steps that need a recorded linear state first, or undo the step in Image History	process_utils::recordNonlinearResult
Apply fails with an error naming the memory budget	The session's memory budget cannot hold the run's copy of the target and the recorded result	Close other images, work on a smaller image, or raise the memory limit on the Processing settings page (it takes effect at the next start); the target is unchanged	process_utils::recordNonlinearResult; ImageRTPreviewControl::finishRun
A cancelled sequence does not replace the target	The sequence run writes a temporary sequence and publishes it only when every frame succeeded; Cancel stops the run and leaves no output	This is expected; restart Apply when ready	ImageRTPreviewControl::runOverSequence
Apply resets the controls	Shared successful-commit path calls Exposure Reset	This is expected; re-enter settings for another still	resetSettingsAsync; ExposureCorrectionProcess::resetSettings

Related workflow steps

Complete calibration and linear color work before using an exposure-like correction. Use Image Statistics and the histogram to measure highlight headroom. Compare Gamma Correction for nonlinear midtone shaping, Histogram Transformation or Statistical Stretch for a main stretch, and Adjust Brightness/Contrast for an encoded-value brightness/contrast control. Monitor Image History after Apply.

References and provenance

• AfterNight project source — authoritative for the current piecewise multiplier, clamp, controls, channel conversion, masks, metadata/state behavior, history, display-only execution autostretch, and the temporary replacement-sequence workflow.
• The standard photographic stop relation m=2^s is used only as the comparison definition needed to interpret the current UI; AfterNight does not implement that equation here.