A growing process-first reference. 25 methods published now, with explicit sources, equivalence, measurements, and validation history; additional chapters will appear when their evidence is ready.
Methodology, measurement, evidence, and shared vocabulary.
Calibration is an upstream operation on raw sensor frames, not a correction applied to a finished stack. Understanding what bias, dark, and flat calibration each remove -- and what clipping and demosaicing destroy permanently -- is what makes later processing decisions interpretable rather than cosmetic.
revision 1 · 14 claimsThe conceptual model behind every candidate comparison NOVA runs: what counts as a fair comparison, what a 'winner' actually means, and why a single run's result stays local evidence rather than becoming a universal recommendation.
revision 1 · 6 claimsHow NOVA judges whether a final image is technically sound, whether processing preserved its information, and which acceptable rendering is preferred -- without turning those different questions into one score.
revision 1 · 9 claimsWhat NOVA actually knows, how it knows it, and how strongly a reader should trust each claim -- the shared vocabulary every other Handbook page draws on instead of re-explaining its own.
revision 1 · 7 claimsNOVA keeps four different kinds of content -- general Handbook guidance, historical Recipe records, claim-specific Validation events, and comparative Experiment evidence -- clearly related but never substituted for one another.
revision 1 · 9 claimsA file's container, name, and appearance are not enough to decide how to process it. A valid processing decision depends on acquisition provenance, spectral response, calibration, geometry, registration, and whether the data are still linear.
revision 1 · 9 claimsHow NOVA's measurements fit into a decision: which numbers are direct statistics, which are proxies, which are heuristics, and where the boundary sits between an objective measurement and an aesthetic preference.
revision 1 · 10 claimsA NOVA result is not reproducible merely because an image file exists. This page separates what must be retained to preserve scientific/process reproducibility from what should be exported for presentation -- those questions often have different answers, and NOVA's own architecture already treats them differently.
revision 1 · 14 claimsA galaxy, an emission nebula, a reflection nebula, a planetary nebula, a globular cluster, and an open cluster can all contain stars, color, gradients, noise, and dynamic range. What actually differs between them is which structures are expensive to damage. NOVA uses object type as one input to adaptive settings and historical priors, but the effective decision still depends on measured noise, PSF, scale, target occupancy, clipping, capture/filter type, and processing state -- a morphology label is a priority and risk layer above the individual process articles, never a substitute for them.
revision 1 · 4 claimsHow NOVA combines Experiment Mode results without hiding failed runs, mistaking repeated targets for independent evidence, or pooling across material version changes.
revision 1 · 6 claimsCalibrate and inspect incoming subframes before they're trusted to stack.
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.
revision 2 · 4 tool pathsInspect comparable light frames before integration, exclude clearly damaging frames, and preserve as much useful integration time as the evidence supports.
revision 1 · 4 tool pathsAlign frames to a common frame and integrate them into one image.
Place eligible exposures on a common sky grid while retaining an honest record of their overlap, so later integration and crop decisions compare like with like.
revision 1 · 4 tool pathsCombine a declared set of registered exposures so persistent astronomical signal accumulates while normalization, weighting, and pixel-level outlier rejection limit noise and transient defects.
revision 1 · 4 tool pathsRefine the stacked image: color, structure, noise, and stretch.
Decide which registered pixels are worth keeping and what sky composition becomes the target's repeatable frame -- a constrained framing decision, not an area-maximization or generic image-quality problem.
revision 1 · 4 tool pathsIdentify and replace isolated defective pixels or very small residual defects without treating stars, compact structure, or ordinary noise as defects.
revision 2 · 4 tool pathsModel and remove unwanted large-scale background variation while preserving faint astronomical signal that occupies the same image.
revision 4 · 10 tool pathsDerive reproducible channel scaling from catalog spectra, the imaging system response, and an explicit white reference instead of balancing color by eye.
revision 2 · 4 tool pathsRecover spatial detail blurred by seeing and optics on linear data, before stretch makes the blur visually obvious and harder to separate from noise.
revision 2 · 4 tool pathsReduce photon and read noise on linear data before stretch amplifies it, without erasing the faint real structure that lives in the same tonal range as the noise. This is Linear Denoise specifically -- NOVA also runs a separate post-stretch noise_reduction pass later in the pipeline; that step changes the artifact and measurement problem (nonlinear data, different noise character) and is not yet its own published Handbook article. Do not treat guidance here as covering both.
revision 2 · 4 tool pathsCorrect star shape -- roundness and elongation, including the FWHM inflation denoise introduces -- without sharpening or deblurring the background or nebulosity.
revision 2 · 4 tool pathsSeparate stars from extended target structure while the image is still linear, so each can receive a stretch and finishing treatment tailored to its own tonal needs, then recombine them into one nonlinear image -- rather than forcing one stretch curve to serve both a star's high-contrast compact profile and a nebula's or galaxy's faint diffuse structure.
revision 2 · 4 tool pathsConvert linear data -- where almost all real signal sits near zero and is invisible on a normal display -- into a non-linear image a human can actually see, without crushing the sky, clipping stars, or hiding the shape of the stretch curve's own tradeoffs.
revision 4 · 8 tool pathsOptionally and temporarily separate stellar from non-stellar content on an already-nonlinear (post-stretch) image, so extended structure -- nebulosity, galaxy detail, dust -- can be finished without simultaneously pushing, recoloring, or clipping stars.
revision 1 · 4 tool pathsCorrect a measured residual sky color cast -- a channel imbalance in a chosen background reference -- after the main color-calibration and stretch path, without modeling a spatial gradient and without acting as a substitute for physical/catalog color calibration.
revision 1 · 4 tool pathsReduce a residual, spatially systematic green excess in the sky/background of an otherwise successfully calibrated and stretched image, without recoloring legitimate green, cyan, or teal target structure -- the SPCC-safe alternative to running broad SCNR everywhere.
revision 1 · 4 tool pathsRefine where already-visible tonal zones sit and how strongly they separate -- shadow/midtone/highlight placement -- after the main stretch has already made the faint signal visible. Curves are final or near-final tonal placement, not a substitute for the main stretch and not a way to recover clipped or never-captured signal.
revision 1 · 4 tool pathsControlled nonlinear chroma enhancement over color relationships that have already been established by calibration -- a rendering transformation applied to calibrated relationships, not additional calibration and not creation of new spectral information. Covers three related but non-equivalent strategies: global saturation enhancement, hue-selective enhancement, and masked enhancement.
revision 1 · 4 tool pathsIncrease tonal separation between nearby structures so real intermediate-scale detail that survived the main stretch, but remains visually compressed against its immediate surroundings, becomes easier to see. This is a finishing operation over already-captured tonal relationships, not a way to restore blur (deconvolution/sharpening) or recover clipped/never-captured signal.
revision 1 · 4 tool pathsRedistribute already-captured nonlinear brightness relationships so bright structures occupy less display range and lower-contrast structure becomes simultaneously visible. This is a tone-relationship operation, not a way to create missing highlight information -- once a region is genuinely clipped in the underlying data or destroyed by an earlier stretch, no HDR operator can reconstruct that lost signal by compressing tone. This article covers the global/multiscale family: NOVA/SASpro WaveScale HDR, PixInsight HDRMultiscaleTransform (HDRMT), and Uri Darom's iHDR. Bright-core selective HDR (`hdr_core_blend`) is a separate strategy that changes spatial treatment scope and is covered by its own article.
revision 1 · 4 tool pathsWhen only a small bright region needs dynamic-range compression, process that region selectively instead of applying HDR to the whole frame. `hdr_core_blend` is not a fourth HDR engine -- it computes a WaveScale HDR layer (the same engine `hdr_compression` uses) and blends it into the frame only under a feathered bright-core luminance mask, typically 0.3-2.7% of the frame. It is a spatial treatment strategy built on an HDR engine, not an independent algorithm.
revision 1 · 4 tool pathsSuppress visually objectionable residual noise -- stochastic grain, chroma speckle, correlated resampling texture -- after the main stretch, while preserving real low-contrast structure and compact features. A nonlinear stretch changes the apparent amplitude and spatial distribution of noise that linear-stage denoise (a separate article) already addressed once; this is an optional second pass on a different signal/noise regime, not a repeat of the same step.
revision 1 · 4 tool pathsSelectively increase the visibility/contrast of dark structures embedded in brighter surroundings -- galaxy dust lanes, dark filaments, faint outer structure -- via multiscale local-contrast redistribution. This is contrast redistribution over existing recorded structure, not recovery of missing signal, not background extraction, deconvolution, denoise, or HDR compression, and not the same operation as PixInsight LHE, CLAHE, or curves, even though all pursue related-sounding perceptual goals.
revision 1 · 4 tool pathsOptionally reduce visually distracting bright-star halo/bloat and color rings after the image is nonlinear. This is a cosmetic/artifact-mitigation finishing transform, not PSF recovery, deconvolution, or star correction -- and the correct action is frequently a no-op. It is a separate concept from dark-structure enhancement (a different article): different objective, different spatial target, different algorithm, different failure modes.
revision 1 · 4 tool pathsAnswer two separate questions that narrowband/dual-band color work conflates if not taught explicitly: what emission-line signal do we think is actually present (the channel/signal layer), and how should that signal be mapped into display RGB (the palette layer)? NOVA has four related but non-alias strategies -- `nb_palette`, `narrowband_norm`, `narrowband_hoo`, and `narrowband_composite` -- that answer these two questions differently depending on what channels are actually available.
revision 1 · 4 tool paths