SubframeQualityReview


Review WBPP subframes, compare their quality, and move rejected original lights after reviewing the selection. User guide for version 0.55.3. [more]

Keywords: WBPP, subframe quality, review, blink, FWHM, eccentricity, PSF signal weight, sky background, autofocus, PHD2, guiding, original lights

Contents

[hide]

1 Description

[hide]

Subframe Quality Review helps you decide which original light frames to exclude before rerunning WeightedBatchPreprocessing (WBPP). It measures the calibrated and registered images from a completed WBPP run, identifies potentially poor subframes within each comparable group, and lets you inspect them against the best scoring subframe or the linear filter master.

The default shortlist contains the worst 10% of measured subframes in each group. REVIEW is a suggestion to inspect an image. Only your explicit rejection decisions are included in a move operation. The corresponding original, uncalibrated files are moved to your chosen folder after you confirm the move plan.

The script supports monochrome images and registered, debayered OSC images from one telescope setup. Optional corner measurements help reveal focus or star-shape problems near the edges. Quality plots can show the acquisition sequence and autofocus start times from the Autorun logs in a selected folder.

2 Start and complete a review

[hide]

With TGScripts installed through the standard PixInsight script installation, choose Script > TG Scripts > Subframe Quality Review. The script requires PixInsight 1.9.6 or later with the ImageAnalysis process installed. Use the documentation button at the lower right to reopen this guide.

  1. Complete a WBPP run. Keep its calibrated and registered lights and its linear, per-filter integrations. Use masters with the same image dimensions as the registered lights; do not crop or drizzle the masters for this review.
  2. Choose one input method: leave WBPP output empty and click Import WBPP, or select the WBPP output and original-light folders manually and click Scan files. Both buttons are in Input files and observing conditions, immediately left of the Autofocus times clock symbol. Enter the telescope aperture for FWHM grading, a destination for rejected originals if needed, and an optional acquisition-log folder.
  3. Check the Filter / group selector and the Original association column. Each original that can be moved must have a unique, verified association.
  4. Set Review worst (%) and the quality weights as needed.
  5. Click the magnifier button at the bottom for a field measurement. To include the center and four corners, click the adjacent Measure corners grid button. It reuses existing field results; if these are missing, it measures the field first. Both actions cover all discovered groups and their matched masters.
  6. Review the measurements and quality plots. Select a subframe and open the Blink frame / reference tab, or double-click its row. The image preview loads when needed. Repeating a measurement reuses unchanged results and an unchanged loaded preview. Inspect the images before deciding which originals to reject.
  7. Optionally click Recommend rejection to compare retained Score and SNR for the selected group. Choose a minimum retained count and, if desired, an object ROI. Analyze does not mark frames; inspect its results before applying the recommendation.
  8. Use the Reject column, the Space key on a selected row, or Reject original in the blink tab to mark unwanted originals. REVIEW flags alone do not mark a file for moving.
  9. Click the green checkmark to review the move plan. Confirm it to move the marked original lights to the rejection destination.
  10. Rerun WBPP with the remaining originals. If you imported a WBPP setup, use the document-export symbol to create a reviewed instance with rejected lights excluded. Otherwise refresh WBPP's light-frame input list manually. Keep the rejection destination out of the new input selection.

Corner measurements automatically adapt parallel processing to the available CPU and memory. No additional settings are required.

Measurements and pending rejection decisions remain in memory while the dialog is open. Closing the dialog discards them. The New Instance button saves folder and evaluation settings in a process icon; it does not save a measured review session. Loading saved settings starts with no corner measurements.

3 Import a WBPP setup

[hide]

With WBPP output empty, click Import WBPP beside Scan files in Input files and observing conditions. Loaded WBPP process icons appear automatically in the WBPP instance selector. Click Open .xpsm to add setups from a saved process-icon file to the same selector. You can open more than one file; loaded icons and entries from other files remain available. Entries identify their source, and hovering over the selector shows the full source path. Opening the same file again refreshes its entries without adding duplicates. Choose the setup for the completed run, check the output folder and number of enabled lights, then click Import.

Import fills WBPP output and Original-light root and populates the review list automatically. It includes only enabled original light frames from the saved setup. Duplicate entries in WBPP groups are counted once. Calibration frames and disabled lights are excluded. The original-light root is the common parent folder of the imported lights.

An instance saved after processing can provide exact registered-image and master paths. Otherwise the script locates the outputs in the WBPP output folder. Verified saved registered-image links remain preferred when other lights need discovery, so older copies in a reused output folder do not replace them. Each filter group also keeps its saved master preference when another group needs a master to be located. Original-light folders are not searched recursively during an import. Files and image headers still have to be checked: saving a WBPP instance does not establish that processing completed successfully. Missing outputs and unverified originals are reported in Messages. Review remains available for verified outputs; an unverified original cannot be moved.

Import replaces the current review, measurements and rejection selections after the files have been checked successfully. Canceling the selector, an invalid setup, or an aborted file check leaves the existing review intact. Aperture, seeing, SQM, log folder, rejection destination and scoring settings are retained. Selected guiding and autofocus logs are refreshed against the imported images, including their inferred UTC offsets. Image measurements are still made with the magnifier and Measure corners buttons.

Scan files is disabled while an imported WBPP list is selected. Import WBPP remains available: choose Saved WBPP setup to recheck that list, or select a loaded icon or .xpsm file to import an updated setup. New Instance retains the imported list with the folder settings; Saved WBPP setup is also available after reopening those settings. Editing WBPP output or Original-light root, or choosing either folder with its folder button, clears the imported list and selects folder scanning. Choosing the same folder also switches to scanning; leaving an unchanged text field does not. When a WBPP output folder has been entered manually, Import WBPP is disabled; clear WBPP output to make Import WBPP available again.

WBPP 2.x through 3.2.x setups are supported. Select the WBPP script icon itself; the ProcessContainer saved in WBPP's logs is a history of processing operations and is not a saved WBPP setup. The imported lights must share a common folder below the drive root and belong to one camera/setup. Existing original-file verification and master compatibility requirements also apply to imports.

4 Create a reviewed WBPP instance

[hide]

After importing a WBPP setup and making rejection decisions, click the document-export symbol immediately left of the green move checkmark. It creates a separate WBPP process icon with the rejected originals removed from all imported light groups. Already moved originals are excluded too. REVIEW flags alone do not exclude a light.

Check the retained and rejected counts, then save the setup as a new .xpsm file. Calibration frames, disabled lights, processing settings and output folder are preserved. Imported lights that are not present in the current review remain in the setup; their count is reported. If a registration or local-normalization reference points to a rejected image, that reference selection is cleared so WBPP can choose again. The imported icon and source .xpsm are not changed.

After writing the .xpsm file to the location you choose, the script launches the installed PixInsight executable, identified by CoreApplication.filePath. It passes the documented --yield option with the current PixInsight instance number and the saved file path. This asks the running instance to load the exported process icon. Close Subframe Quality Review to let PixInsight load the new WBPP icon. Double-click that icon to inspect the setup before running WBPP. The new icon uses WBPP from the current PixInsight installation. The script does not start processing automatically. The saved .xpsm can also be loaded manually with Process Icons > Load. Existing process icons are preserved.

Also move rejected originals is optional and initially off. With a Rejected originals folder selected and pending moves available, enable it to combine creation with the existing move workflow. A separate move confirmation lists the files and destination. The setup is saved before moving; canceling the move leaves it available. Failed moves retain the existing journal and undo protections. Creating the instance alone does not move any image files.

The button requires a successful WBPP import with complete settings and at least one rejected or moved original. If its tooltip asks for another import, select the original loaded WBPP icon or .xpsm again. All rejected originals must have verified associations, at least one enabled light must remain, and retained light files must exist. Manually selected folders do not provide the complete WBPP setup needed for this action.

The new instance is a snapshot of the decisions when it was created. If you later keep a rejected frame again or undo a move, create another instance to reflect that change. Choose an appropriate output folder in WBPP before starting the new run.

5 Input files and observing conditions

[hide]

Drag a window edge or corner to adjust its height and width. The minimum width keeps the Quality evaluation groups on one row, with compact Guiding controls and unwrapped threshold text. Folder paths are sized to this minimum width and stay fixed when the window is widened. The Autofocus times clock symbol remains aligned with the right edge of the path text fields, with Scan files and Import WBPP immediately to its left. The compact Filter / group selector and Quality evaluation groups also keep their widths; the review table and previews use the additional space. Some preview controls can move to additional rows as the window narrows. The window scrolls vertically only; the Review table has its own horizontal scrollbar for additional columns. Hover over Filter / group for the full group name. The log summary and Autofocus times stay on the UTC offset row even at minimum width; longer log summaries wrap within their text area. The available input action is highlighted when the review list needs to be populated. The REVIEW filter stays inside Sort and filter. The action buttons remain visible at the bottom. Workflow guidance and current messages appear together below the tabs.

The Filter / group label aligns with the folder labels, with its selector immediately beside it. The SQM time Choose button shares the right edge of Autofocus times and the path text fields. In Guiding, the threshold message sits directly below Weight, aligned with the text of Guiding in Score.

Click the Input files and observing conditions bar to collapse or expand the telescope, folder, and log controls. It opens expanded by default. The window keeps its size; collapsing a section reduces scrolling or gives the active tab more room. Collapsing it retains the selected paths, settings, measurements, and rejection choices. Saved process instances remember its expanded or collapsed state.

Telescope aperture is at the bottom of this section, aligned with the folder controls. The aperture, seeing, Session SQM, and reading time have evenly spaced control groups; the reading time moves to a second row only when the controls no longer fit together. Enter the imaging telescope's aperture diameter in whole millimeters. It is required for FWHM grading and is initially unset; it is not needed to measure or review the images.

Zenith seeing is beside Aperture. Enter an optional seeing estimate in arcseconds at the zenith and a reference wavelength of 550 nm. A positive value enables grading relative to the expected FWHM for those conditions. Leave it at zero to use the fixed Good and Medium seeing limits. A forecast is an estimate, not a measurement of the actual conditions; check its reference conditions before entering it. If it already includes the target altitude correction, convert it to zenith seeing first.

Session SQM and SQM time (local) are beside Aperture and Zenith seeing. Enter the reading in mag/arcsec². Loading the log folder presets the date and local time from the earliest light-frame sequence in its Autorun logs. Flat frames, focusing and plate-solving exposures, and Chinese translations are ignored. Adjust the preset to the actual SQM reading time. A time you enter manually is retained when the logs are reloaded or changed; an untouched automatic preset follows the selected log folder. This distinction is retained in saved settings. If the new folder has no recognized light-frame start, an automatic preset is cleared.

Click Choose beside SQM time to select the year, month, day, hour, minute and second, then Apply. Cancel retains the existing time. If the time is empty, the selector starts at the logged imaging start, then falls back to the selected exposure midpoint when its local time is known, or the computer clock if neither is available. You can also type or paste YYYY-MM-DD HH:MM or YYYY-MM-DD HH:MM:SS directly. The date avoids ambiguity when a session crosses midnight. Log presets keep their local wall time; the existing inferred or custom UTC offset converts the reading to image UTC timestamps. The reading automatically determines the approximate Bortle class.

Manual entries are checked for a valid calendar date and clock time. On Enter or leaving the field, valid entries are formatted as YYYY-MM-DD HH:MM:SS; omitted seconds become 00. A T between date and time or slashes in the date are accepted and standardized. Do not append a timezone: the UTC offset controls supply it. Invalid entries remain visible with a highlighted field and an explanation in Messages. They cannot be used for calibration. Correct the entry or use Choose; the script does not silently replace it with another time. Committing a valid correction also clears its validation message in Messages, while preserving newer messages from other operations.

Zero means that no SQM reading is supplied. In that case, select Sky brightness on the Quality indicators tab to choose a Bortle class from the compact selector to the left of Calibrate sky. Choose a class from 1 to 9, or — when unspecified. The starting selection is class 6. Your manual selection is retained while SQM determines the class automatically.

Table 1

Input

What to select

WBPP output

The completed output directory containing the registered subfolder and, when available, the master subfolder.

Original-light root

A directory containing the original, uncalibrated light files. Subdirectories are scanned recursively. Flats, darks, bias frames, and master files are excluded from original-light matching. The original root must be outside the WBPP output directory.

Log folder

An optional folder containing Autorun_log and PHD2_GuideLog files for the acquisition sessions. Files marked _CHN are ignored.

Rejected originals

The destination for files you manually reject. Relative subdirectories from the original-light root are preserved. This folder is excluded from original-light scanning.

XISF and FITS light images are supported. Groups follow the WBPP registered-image directories and compatible exposure times, dimensions, and channel counts. Different filters or exposure groups are evaluated independently. Scores and ranks describe relative quality within a group; they are not absolute quality ratings across filters or sessions.

The master is selected automatically from the master subfolder. Its masterLight filename must identify the group's filter. When several masters share that filter, the script looks for the corresponding WBPP group name. A missing or ambiguous match is reported instead of choosing an arbitrary master. The best scoring subframe remains available as a reference when no master is matched.

The bottom message area shows one shared pixel scale in arcseconds per pixel and the camera name, independent of the selected filter. The scale comes from the master metadata's focal length and effective pixel size; X and Y values appear separately when they differ. Hover over the scale to see the master filenames and calculation inputs. Scan again after changing a master file.

The registered images, their associated originals, and matched masters must identify the same camera. A scan stops if different cameras are found. Missing camera names are reported as not fully verified. The same camera can still have a different pixel scale with different focal lengths or binning. Conflicting master scales or different master and registered-image dimensions are reported instead of assuming one shared scale. If no usable master scale metadata is available, the scale is shown as unavailable.

Original matching uses the source filename and acquisition metadata, including observation time, camera, and exposure. An ambiguous match, missing required metadata, or an original associated with more than one processed image prevents that original from being moved. The Original association column and row details explain the result. Master rows are for comparison and can never be rejected.

Changing the WBPP output or original-light root invalidates the current review; scan and measure the new selection.

Scan files and Import WBPP are alternative input methods. Scan files becomes available when both manual image folders are specified; Import WBPP is available when WBPP output is empty or an imported setup is selected. The available action is highlighted in slate gray, retaining the standard pushbutton appearance, when the review list needs to be populated. The highlight clears after a successful scan or import. Hover over either button for guidance on switching methods. The clock symbol opens Autofocus times; it is available when a log folder is selected. The magnifier, Measure corners and Undo moved originals buttons are enabled only after a successful scan or import. Changing measurement or display settings does not require another file scan.

The green Move checkmark and blue Undo arrow also require a Rejected originals folder. Both remain disabled while this field is blank. Their tooltips explain what is required. Entering or clearing the folder updates the buttons immediately; no rescan is needed. Move additionally requires at least one manually rejected original. Scanning and measuring remain available with the rejection folder blank.

6 Quality evaluation controls

[hide]

Click the Quality evaluation bar to collapse or expand the section. It opens expanded by default. The window keeps its size when either section is toggled. Collapsing it retains all settings, measurements, scores, and rejection choices. Saved process instances remember the expanded or collapsed state.

The first column separates Measurements and review from Score selection. Measurements and review has Review worst and Minimum stars in the left column, with Edge guard and Corner size in the right column. Score selection contains the three checkboxes that determine the active Score, with the bold Current Score message to their right identifying the selected R1, R2, R3, or R4 method. Score weights sets the manual weights in two columns. Guiding contains its scoring mode, weight, and thresholds. The active scoring mode is also reported in the bottom message area.

With the R3 and R4 options off, manual scoring uses R1 for field-only evaluation or R2 when corner measurements are included. The strongly correlated KPI option selects R3; the rank-of-ranks option selects R4. These two options are alternatives. The corner checkbox determines whether available corner measurements participate in the selected method.

Table 2

Control

Default

Meaning

Use R3: strongly correlated KPIs

Off

Use the R3 correlation score as the active Score. R3 is also calculated and displayed while this option is off. Selecting it turns off Use R4: rank of ranks.

Use R4: rank of ranks

Off

Use the equal-weight average of normalized R1, R2, and R3 ranks as Score. Omit unavailable ranks and omit R2 while corner scoring is off. Selecting it turns off Use R3: strongly correlated KPIs.

Review worst (%)

10

Shortlist this percentage of valid field measurements independently in each group, rounded up. Images with insufficient required measurements are also flagged for inspection, so REVIEW may contain more than the percentage quota.

Measure corners

Button beside the magnifier

Measure the center and four corners across all groups and matched masters. Existing field and regional results are reused. Missing field results are measured first. Regional results appear in the selected-image measurements; the scoring checkbox controls whether to use them.

Include corner measurements (R2)

Off until measured

Include or exclude stored corner results in the current score without measuring again. Available once regional results exist. The first corner measurement enables it; subsequent measurement runs retain your choice.

Corner size (%)

22

Width and height of each corner and center patch, as a percentage of the common measurement field.

Edge guard (%)

2

Inset applied inside the common valid field to reduce edge contamination.

Minimum stars

10

Minimum star and PSF support required for a valid measurement. Sparse fields or small corner patches may fail this requirement.

Quality weights

See the Score section

Relative influence of each component. A zero weight disables that component. In manual mode, at least one underlying quality indicator must have a positive weight.

Sky KPI

Sky quality, weight 20

Choose the single sky indicator eligible for scoring: Sky quality, Transparency, or Sky brightness. Sky quality combines transparency and sky background. The other two remain available for inspection.

The Sky KPI selector in Score weights applies to R1, R2, Below-mean Quality, and the sky candidate considered by R3. R4 uses the resulting ranks. Only the selected sky indicator can participate; R3 does not substitute another sky indicator if your choice fails its correlation threshold. The shared weight controls manual scoring; R3 uses weight 1 when the selected indicator qualifies. A zero manual weight excludes it from R1, R2 and Below-mean Quality, while R3 still tests the selected indicator as it does other candidates. The selection and manual weight are saved with process settings. Changing the selection recalculates scores from existing measurements without remeasuring images.

Changing the review percentage or quality weights recalculates rankings from existing measurements. Changing Minimum stars immediately reassesses the measured star counts and updates validity and scores; stored field and corner measurements are retained, and no images are remeasured. Measure corners runs the missing regional measurements and updates scores and correlation selection. It does not repeat existing full-image PSF measurements while the images and measurement settings are unchanged. An unsuccessful regional result remains marked as unavailable; clicking the button again does not repeatedly retry it. Changing Corner size or Edge guard clears the measurements and resets the corner status. Click Measure corners to measure the previews with the new settings, or use the magnifier for field measurements.

Include corner measurements (R2) switches between field-only and field-plus-corner evaluation. Turning it off excludes corner penalties and individual corner KPIs, uses the whole-image sky background, and recalculates Score, Below-mean Quality, R3, and R4. Turning it on uses the stored regional results again. Neither action remeasures images. R1 and R2 remain available for comparison, and the regional measurements and plots are retained. Missing results are still reported when their scoring contribution is enabled.

If regional measurements are already available, their quality plots remain available when corner scoring is turned off. The plot indicates that those measurements are excluded from the current Score. Corner-specific choices are hidden when no corresponding measurements exist. In the Correlation plot, only measurements from the current group's subframes enable regional choices; master-only measurements do not.

7 What is measured

[hide]

The script calls PixInsight's ImageAnalysis process to measure the registered lights and the matching masters. It uses a shared valid field inside the registered images, excluding registration borders as far as the sampled border checks can determine. FWHM is reported in pixels with one decimal place.

Measurements are performed on temporary, floating-point image crops. This allows signal and noise information to be evaluated for the measured region without reusing inherited WBPP noise metadata. Source images are opened for reading; their pixels and metadata are not changed.

Field measurements cover all groups and their matched masters. Optional center and corner measurements follow, with parallel processing adapted to your computer's available memory and PixInsight processing limits. Existing field measurements are reused when adding corners.

After you click the magnifier, the script checks the common usable field, measures missing field results and any enabled missing regions, and updates the scores. For sky calibration, an SQM reading uses the measured frame nearest its reading time. Without SQM, a selected Bortle class uses Gaia photometry on the best default-scoring frame. Gaia sky-coordinate repair, when needed, can make this last step take longer. The console shows progress. Repeating measurement in the same open review reuses unchanged results and sky calibration. Opening the Blink tab loads the image preview when needed.

With corner scoring enabled, each of the four corners is evaluated for FWHM, FWHM relative to the center, and eccentricity. These measurements use the registered image orientation. After rotation or a meridian flip, a displayed corner does not necessarily correspond to the same physical camera corner.

Measurements describe the processed, registered images. Their PSF signal weights need not equal the values originally recorded by WBPP at an earlier processing stage. Very sparse regions, unusual star profiles, or registration artifacts merit visual inspection even when a numerical score is available.

8 Rejection recommendation

[hide]

After field measurements, select a Filter / group and click Recommend rejection beside the measurement buttons. This compares trial stacks to suggest how many of the remaining frames to retain. Existing rejected or moved frames and the integrated master are excluded. Each filter and exposure group is evaluated separately. It uses the current Review Score without changing Score, R1–R4 or their weights.

Preserve SNR (%) defaults to 95. The recommendation must retain at least this percentage of the highest SNR measured among the tested stacks. Increase it to favour depth, or decrease it to allow more loss of depth in exchange for a higher mean retained Score. Minimum retained defaults to 20 frames; choose a larger value when you want more statistical support. Seven is the lowest permitted value for the trial's pixel rejection, not a guarantee of reliable rejection under every condition. If too few frames remain, the script does not make a recommendation.

Without Use object ROI, the script evaluates the common usable field. SNR is a stellar proxy and may not represent the faint structures of your target. With Use object ROI, enter X, Y, Width and Height in registered-image pixels, or select an existing PixInsight preview on an aligned image of the same dimensions. The ROI must lie within the common usable field and have enough surrounding sky. It must be at least 32 × 32 pixels.

Include the target structures whose signal you want to preserve. A galaxy's faint outskirts may favour a different selection from its bright core. Avoid unrelated bright stars inside the rectangle: all aperture content contributes to its signal. Leave clean sky outside it for background estimation. A guard band separates the sky samples from the target, and bright stars in the surrounding sky are masked. The same object aperture and sky mask are used throughout the comparison. If surrounding sky is insufficient, make the ROI smaller or reposition it. PSF measurements require enough stars in the common usable field, independently of the object ROI. A saved ROI is reset when the selected group has different image dimensions.

Click Analyze to prepare and test the candidate stacks. The current Review Score orders the remaining frames from best to worst, independently of the table sort. Each trial removes frames from the lowest-scoring end. The chosen R1, R2, R3 or R4 score mode applies; select R3 in Review if you want its strongly correlated KPI score. Every remaining frame needs a valid field measurement and Score. Scores and their ordering stay fixed throughout the comparison. Equal scores are ordered by filename.

The plot shows retained frame count on the horizontal axis and preserved SNR in blue. Use Plot KPI to choose the orange curve: Score (retained mean) is the default. Other choices are FWHM, PSF signal weight, PSF SNR, Stars, Eccentricity, Noise (sigma), Retained exposure, and Effective frame count. A previously saved plot selection is restored. The selected KPI also appears in the table. Changing this selection redraws the existing measurements immediately; it does not run new integrations or change the recommended cutoff. The recommendation always uses mean retained Score, regardless of the plotted KPI.

Score (retained mean) is the arithmetic mean of the retained subframes' Review scores at analysis time. It does not assign a Score to the integrated image. Changing scores, weights or the scoring mode requires new analysis because it can change the candidate ordering and recommended cutoff. PSF signal weight, PSF SNR, Stars, Eccentricity and Noise are measured on each trial stack. PSF SNR is shown as an amplitude ratio. Exposure and effective frame count describe the retained set. Hover over the selector for the selected indicator's definition.

The object ROI affects only Object SNR. Its background is estimated from nearby sky outside the aperture. FWHM, PSF signal weight, PSF SNR, Stars, Eccentricity and Noise always use the common usable field. Trial-stack normalization also uses this field, independently of the ROI. For the same retained frames, changing the ROI therefore leaves those image KPIs unchanged. It can change which counts meet the SNR limit and which additional counts are tested. The area is identified beside Plot KPI. Score, exposure and effective count describe the retained frames, independently of the ROI.

Stars counts detections made independently in each trial stack; it does not track a fixed list of matched stars. Changes in noise, star profiles and pixel rejection can move marginal detections across the detector's acceptance criteria. Counts can therefore rise or fall between neighbouring trial counts, especially in a small field. This does not mean that the measurement area changed, and a larger count alone does not establish a better stack.

The dashed blue line is your SNR preservation limit; the green line is the recommended count. Hover over a plotted point for its values, including the unrounded selected KPI. The table reports tested counts, proposed rejects, SNR retention, the selected KPI and exposure time. Hover over a row for its effective weighted frame count, measured star count, or failure reason. Missing KPI measurements remain unavailable rather than appearing as zero.

The recommended count gives the highest mean retained Score among tested stacks that satisfy the SNR and minimum-count limits. Equal mean scores favour more retained frames. FWHM does not determine the rejection decision. The all-frame baseline is tested too. The search tests several cutoffs and neighbouring counts around the best result; it does not guarantee a global optimum over every possible subset. Failed trials remain visible. If the all-frame baseline fails, no recommendation is issued.

Object SNR is a background-limited aperture estimate above the surrounding sky, with allowance for nearby correlated pixels and uncertainty in the background estimate. A faint object's measured signal may be zero or negative in individual subframes; these frames remain eligible. Each successful trial stack must have positive background-subtracted object signal. The estimate does not include source shot noise, shared calibration errors or long-range correlations. Nebulosity filling the surrounding sky, gradients not represented by a plane, and foreground stars inside the aperture can bias it. For bright targets or very faint outskirts, inspect the images and treat small differences cautiously. Without an ROI, the native PSF power ratio is converted to an amplitude ratio so the percentage limit has the same interpretation.

Trial stacks use fixed PSF signal weights, global photometric normalization, and Winsorized pixel rejection. The test rejection limits are 4 sigma low and 3 sigma high. RGB comparisons use green. Local normalization and drizzle are not reproduced; different integration settings can change the eventual outcome. Temporary crops are reused across tests and removed afterwards. All tests integrate the common usable field, including when Object ROI is enabled. Analyze can take time for large images; Stop cancels it without marking frames.

Mark recommended rejects adds rejection marks for the proposed frames and completes REVIEW for the analyzed group. Its retained frames no longer remain on the REVIEW shortlist. Marking all proposed rejects manually has the same effect. When the recommendation retains every frame, Complete review accepts that decision without rejecting any. Inputs and settings are checked again before applying; if they have changed since Analyze, no decisions are changed and a fresh analysis is required. Existing rejection decisions, measurements and scores are preserved; no files are moved. Inspect the marked frames in Review or Blink; the usual green checkmark opens the separate move plan. Closing without applying leaves decisions unchanged.

Review completion remains valid for the analyzed settings and measurements. Running Analyze again with unchanged inputs creates a new proposal without undoing a completed review. Changing settings, measurements or other keep/reject decisions reopens review; keeping a proposed rejected frame again also reopens it. Moving an already rejected original does not reopen review. The REVIEW-only filter is empty when review is complete.

Closing the recommendation window retains its results and settings for each group while the main review window remains open. Click Recommend rejection again to display them without reading image pixels or repeating the integrations. An applied recommendation can also be reopened; it is identified as already applied. Moving its marked originals preserves these results while the registered inputs remain unchanged. An interrupted refresh with unchanged inputs retains the previous completed results.

Results are cleared when their analysis settings or inputs change: for example, the SNR limit, minimum retained count, active ROI, relevant measurement settings, scores, scoring weights or mode, corner/guiding participation, selected sky indicator, measurements, source files, or manual rejection decisions. Plot selection, Review sorting and display settings retain the native results. Coordinates of a disabled ROI do not affect them. Each group keeps its own recommendation settings, so visiting another group does not discard its results.

Closing the main review window or scanning a new session discards the measured session and its cached recommendations. Saved script settings retain the last SNR limit, minimum count, optional ROI and plot selection; they do not save trial results or rejection marks.

9 Sky brightness

[hide]

With Include corner measurements (R2) enabled, Sky brightness is the minimum of the arithmetic means of the five previews: Center, Top left, Top right, Bottom left, and Bottom right. With the option off, it is the arithmetic mean of the whole registered image, including any registration borders.

Values are reported in image units. For RGB images, the mean uses the green channel. Stars and extended objects remain part of the means; this indicator is an estimate, not a measurement of empty sky alone. Compare images with the same brightness scale and processing. Background subtraction or normalization can change its interpretation.

Lower values are treated as better quality. To use this indicator in scoring, choose Sky brightness in the Sky KPI selector. R1 uses the whole-image mean; R2 uses the preview minimum. Only this selected sky indicator then participates in Below-mean Quality and is eligible for R3 selection. Switching corner scoring updates the displayed value, Score, R3 selection, and R4.

The Review table reports Sky brightness with six decimal places. Hover over the value to see its meaning and the five preview means when available. Choose Sky brightness on the Quality indicators or Correlation plot tab, or in Sort by. Missing background measurements are shown as unavailable and do not count as good quality. All five preview means are required to establish their minimum; star-detection success is not required for a pixel mean.

10 Transparency

[hide]

Transparency uses the RelativeExtinction measurement from ImageAnalysis to compare stellar light within each filter group. Higher percentages indicate more transmitted stellar light. Measurements account for exposure time. For RGB images, ImageAnalysis uses HSI intensity.

The most transparent measured subframe supplies the 100% reference. Transparency (%) = 100 × 10^(−0.4 × relative extinction in magnitudes). An extinction of about 0.242 magnitudes corresponds to 80% transparency. This is a relative comparison, not an absolute atmospheric transmission measurement.

Use the magnifier or Measure corners button to obtain the measurement. The Review table reports the percentage; its tooltip identifies the reference and relative extinction in magnitudes. Transparency is also available in Sort by, on the Quality indicators tab, and on either axis of Correlation plot. The plot information button explains the measurement and identifies the group's reference.

ImageAnalysis requires at least two comparable subframes with valid exposure times and enough usable unsaturated stars in a common brightness range. If these conditions are not met, Transparency is unavailable. A valid zero extinction means 100% transparency; an undefined measurement does not.

The master is compared separately with the same subframe reference and appears in Review and in the quality plot. It never influences the subframes' transparency measurements or determines their reference, and remains excluded from correlation plots. The comparison uses the master's recorded frame exposure, not its total integration time. Without a valid frame exposure or a reliable subframe comparison, the master value is unavailable. Its percentage can exceed 100%.

For the master, this is the effective stellar throughput of the integrated image. Brightness normalization during integration can change it, so it does not represent the average atmospheric transparency of the session. The Review tooltip identifies this distinction.

The reference is independent of Score, the current sort order, rejection flags, and whether corner scoring is enabled. Choose Transparency in the Sky KPI selector to use it in scoring. Higher values are better. Sky brightness and combined Sky quality then remain comparison indicators and do not enter scoring separately.

Clouds can reduce transparency, but changing target altitude, focus, and image brightness normalization can also change the result. Compare calibrated, aligned subframes before brightness normalization; a low value alone does not establish the cause. The measurement is reused when adding corners or changing scoring settings.

11 Cloud extinction

[hide]

Cloud extinction uses ImageAnalysis to estimate stellar dimming beyond the normal change with airmass. It is reported in magnitudes in the Review table, in Sort by, on the Quality indicators tab, and on either axis of Correlation plot. Lower values indicate less additional dimming. It is a comparison indicator and does not enter Score or R1–R4.

The calculation is Cloud extinction = Relative extinction − (a + k × airmass). ImageAnalysis fits the atmospheric baseline robustly within each filter group, reducing the influence of unusually dim frames. Values near zero indicate little additional dimming. Positive residuals can indicate clouds, haze, or dew. Negative residuals are retained; they can occur through scatter around the fitted baseline. This is a relative estimate, not an absolute measurement of cloud cover.

At least five comparable subframes with usable stellar measurements, valid exposure times, and known airmass are required. Airmass is derived from each frame's observation time, target coordinates, and observing location; recorded target altitude can provide a fallback. The manually chosen target altitude for FWHM grading is not used. If the group's airmass range is below 0.05, ImageAnalysis fits only a constant baseline.

Observation times can be stored as complete timestamps or as separate date and time fields. Both standard FITS names (DATE-OBS, TIME-OBS, DATE-END, TIME-END) and their unhyphenated aliases are supported. Valid XISF observation times take priority, followed by standard FITS names and then aliases. Times without an explicit offset are interpreted as UTC; explicit offsets are converted to UTC. Incomplete or invalid timestamps are treated as unavailable, so a date alone cannot be mistaken for a midnight exposure.

Use the magnifier or Measure corners button to obtain the measurement. Hover over its Review cell to see the calculated airmass and altitude, or why the measurement is unavailable. The plot information button explains the calculation and the group's available measurements. An undefined value appears as a dash, not zero.

Integrated masters are not applicable and are omitted from this indicator's plots. Adding corners and changing scoring settings reuse the field measurement.

12 Sky quality

[hide]

Sky quality combines the existing Transparency and Sky brightness indicators. Higher values favour frames with more transmitted stellar light and a darker background. It is available as an additional column in Review, in Sort by, on the Quality indicators tab, and on either axis of Correlation plot.

The combination is Q = (T / Tref) × sqrt(Bref / B), displayed as a percentage. T is Transparency. B is the current Sky brightness value minus any recorded calibration pedestal, divided by exposure time. No separate noise measurement is required. The transparency reference is also the Sky quality reference, at 100%; values above 100% are allowed.

For example, unchanged transparency with twice the sky-background rate gives about 71% Sky quality. A frame with 80% of the reference transparency and half its background rate gives about 113%.

With Include corner measurements (R2) enabled, the background uses the minimum of the five preview means. Otherwise it uses the whole-image mean. Changing this option updates Sky quality immediately from the existing measurements. The reference and calculation details are available in the Review tooltip and the plot information button.

Sky quality is a relative estimate of the transparency/background balance, not an absolute sky-brightness measurement or proof of cloud cover. Its relationship to faint-object signal quality assumes comparable camera settings and pixel scales, unnormalized linear data, and exposures dominated by sky background. Stars and extended objects in the background mean, read noise, image normalization, and changes in focus can limit that interpretation. Strong correlation between its two inputs alone does not establish that they measure the same physical quantity or that this combination is superior.

Both indicators, a valid exposure time, and a positive background after pedestal removal are required. Missing or invalid values appear as unavailable. The master uses the same formula and subframe reference, with its recorded frame exposure. It appears in Review and in the quality plot, while remaining excluded from correlation plots. Its Sky quality describes the integrated image's transparency/background balance and is affected by integration normalization; it does not measure the signal-to-noise improvement from stacking.

Sky quality is the default choice in the Sky KPI selector, with manual weight 20. It combines transparency and sky background; higher values are better. Sky brightness and Transparency supply its calculation but do not also enter scoring as separate indicators. R1 uses the whole-image background basis; R2 uses the preview minimum. It uses cached measurements and adds no image reads or star measurements. Adding corners or changing scoring settings reuses the master's stellar measurements as well as those of the subframes.

13 Score and Below-mean Quality

[hide]

Higher Score means better relative quality. Scores range from 0 to 1 and are displayed with three decimal places. The default Score sort puts the lowest, worst scores first. The score combines penalties for weak signal, few detected stars, large or elongated stars, poor results in the selected sky indicator, optional corner problems, guiding errors when logs are available, and the breadth of below-mean quality across indicators.

Table 3

Component

Poor-quality direction

Default weight

PSF signal weight

Lower

40

Star count

Lower

15

Field FWHM

Higher

25

Field eccentricity

Higher

20

Corner quality

Higher penalty; included when enabled

20

Selected Sky KPI

Lower Sky quality or Transparency; higher Sky brightness

20

Guiding quality

Larger combined guiding error relative to image pixel scale

25

Below-mean Quality

Higher fraction of indicators worse than their mean

20

For each image-based field indicator, the reference is the median of the group's supported subframe measurements. Its scale is the larger of 1.4826 times the median absolute deviation, 2% of the absolute median, and a tiny numerical floor. The master is excluded from these reference statistics.

For an indicator where higher is worse, the penalty is (value minus median) divided by the scale. For an indicator where lower is worse, it is (median minus value) divided by the scale. Negative penalties become zero, and penalties above 10 are capped at 10. A good result in one indicator therefore does not cancel a poor result in another.

Corner quality is the worst penalty for star size, corner-to-center size ratio, or elongation across the four corners. Lower is better.

Below-mean Quality uses arithmetic means, not medians. It is the percentage of assessed, enabled underlying indicators on the poor-quality side of their mean. Low PSF signal weight and low star count count as worse; high FWHM, high eccentricity, high Corner quality penalties, and high guiding error count as worse. The selected Sky KPI counts once: lower Sky quality or Transparency is worse, while higher Sky brightness is worse. Equality with the mean is not worse. Score and Below-mean Quality itself are excluded from this comparison.

For example, a subframe worse than the mean in three of four assessed indicators has Below-mean Quality of 75%. Its score penalty uses the fraction 0.75 directly. With the default weight of 20, it adds 0.75 times 20 to the weighted numerator. Setting the Below-mean Quality weight to zero removes this score contribution while keeping the percentage visible.

Score = 1 − sum(weight × penalty) / (10 × sum of weights). Only available components enter the sums. A Score of 1.000 means no measured penalty; 0.000 means the weighted mean penalty reaches its limit of 10. Score describes relative quality within its group, not a probability or an absolute image-quality rating. Rankings use full precision, so values that display identically can have different ranks.

The arithmetic means used for Below-mean Quality include supported subframes in the group, including manually rejected subframes, and exclude the master. The saved manual weights determine which underlying indicators participate: a zero weight excludes an indicator from this comparison. Corner quality counts as one indicator when corner scoring is enabled and weighted. Selecting R3 or R4 changes the active Score without changing Below-mean Quality, its completeness indicator, or its correlation. Changing the manual KPI selection or the corner/guiding settings can change this composite.

Unavailable indicators are not treated as good measurements. An incomplete Below-mean Quality percentage is marked with an asterisk; its tooltip gives the assessed and enabled counts. A partial Score may still be displayed, but missing required data triggers REVIEW, and R1 and R2 are withheld when their required measurements are unavailable. A partial R3 remains available when the field measurement is valid; inspect its missing-data flags.

R3 is calculated when field measurements are available and at least one KPI qualifies. Enable Use R3: strongly correlated KPIs to use its score as the active Score. It selects indicators automatically for each group when their absolute Pearson correlation with PSF signal weight is at least 0.7, accepting both positive and negative relationships. It requires at least three paired subframes with varying measurements. The master is excluded; manually rejected subframes remain included. Strong correlation describes association, not the cause of poor quality, and can be unstable in small groups.

PSF signal weight is the correlation reference only. It is excluded from R3 selection and contributes no direct weight to R3; self-correlation is not used as selection evidence. Each qualifying indicator has weight 1; excluded indicators have weight 0. Manual weight values are retained, including any zeros, and restored when the option is disabled. Corner quality participates when Include corner measurements (R2) is enabled; guiding also requires logs and a guiding mode other than Ignore. The selected guiding mode still controls missing-data handling when guiding is included.

With Include corner measurements (R2) enabled, the FWHM and eccentricity of each of the four corners are also tested separately against PSF signal weight. Each of these eight indicators is included with weight 1 when its absolute correlation reaches 0.7 with at least three valid paired subframes. The Corner quality composite remains a separate candidate. A corner FWHM can qualify even if eccentricity or another corner is unavailable. Center FWHM contributes through the composite's corner-to-center ratios; it is not a separate scoring candidate.

Each selected corner indicator contributes its own penalty to R3. Larger FWHM and eccentricity are worse, regardless of the correlation's sign. Several qualifying corner indicators therefore give corner performance more combined influence. The Below-mean Quality composite retains its manual indicator selection; qualifying individual corner indicators do not redefine it. Disabling correlation scoring restores the manual weights and the single Corner quality contribution.

R3 assesses the same Below-mean Quality values shown in the Review table and correlation plot, and includes this composite only if its own correlation passes the threshold. Score itself is never a selection candidate. If no KPI qualifies, the script reports No strongly correlated KPIs and leaves the R3 score and rank unavailable. With Use R3 enabled, Score retains the manual evaluation, including its weights, review reasons, and ranking. The weight controls show those retained manual values. R4 omits the unavailable R3 rank. This behavior is determined separately for each filter group.

The qualifying R3 KPIs are listed in the bottom message area on every tab. Hover over that report or R3 for each correlation, sample count, and exclusion reason. Changing this option reuses existing measurements.

Clicking Measure corners afterwards measures missing regions and recalculates the selection for every group. When corner scoring is enabled, Corner quality and the eight individual corner indicators are included only when each meets the strong-correlation threshold. Turning corner scoring off excludes them and recalculates the selection using the field-only indicators. Clearing measurements also removes their scoring contributions until they are measured again.

The master provides comparison measurements but receives no Score or R1–R4 ranks. These cells show an em dash. It is excluded from score baselines, rank populations and the REVIEW quota. Its measured quality indicators remain available in the table, measurement plots and blink preview.

14 R4: Rank of ranks

[hide]

R4 combines three views of image quality into one ranking. Each available ranking has equal influence after normalization. It averages the rank positions, rather than the underlying quality scores.

Table 4

Component

What it evaluates

R1

Field quality using your manual weights, without corner measurements.

R2

Field and corner quality using your manual weights.

R3

Quality using equally weighted indicators that strongly correlate with PSF signal weight.

For each component, normalized rank = (rank − 1) / (N − 1), where N is the number of images with a valid rank for that component in the group. The worst rank becomes 0 and the best becomes 1. If only one image has a rank, its normalized value is 0.5.

R4 Score = mean of the available normalized R1, R2, and R3 ranks.

The R4 column then ranks these combined scores: 1 is worst, and larger rank numbers are better. The combined Score itself ranges from 0 to 1, with higher values indicating better relative quality. Rank and Score are different values.

For example, suppose each component ranks 11 images and one image has R1 = 3, R2 = 7, and R3 = 9. Its normalized ranks are 0.2, 0.6, and 0.8. Its combined Score is (0.2 + 0.6 + 0.8) / 3 = 0.533. Its R4 position depends on the combined scores of the other images. If only R1 and R3 are available with those values, the combined Score is (0.2 + 0.8) / 2 = 0.500.

R2 is omitted when Include corner measurements (R2) is off or that image has no valid R2. Other unavailable ranks are also omitted; they never count as zero. An image with no available ranks has no R4. A partial result can still require REVIEW. Corner scoring also affects which indicators participate in R3, so switching it can change both R3 and R4.

To use this method, enable Use R4: rank of ranks in Score selection. The Score column, REVIEW shortlist, and best-subframe comparison then use the combined score. Use R3: strongly correlated KPIs selects the R3 score instead; the two options are alternatives. Leaving both off uses manual scoring. R1–R4 remain visible in every mode when their required measurements exist, and changing modes does not measure images again.

Select R4: rank of ranks in Sort by to order the table by R4, independently of the active Score mode. Hover over an R4 value to see the contributing ranks, their population sizes, normalized values, and any omissions. R4 uses all ranked subframes in the group regardless of the table filter, excludes the master from each population, and resolves equal scores by filename.

Manual weights remain editable in rank-of-ranks mode because they determine R1 and R2. R3 uses its independently selected, equally weighted indicators. Below-mean Quality has the same definition in every scoring mode. The Main issue and Score tooltip identify the ranks that most reduce the combined score.

15 Review and measurements

[hide]

Above the table, Sort and filter places the sort criterion and order on one row, with the REVIEW filter below. Beside it, Reading the table shows the rank descriptions on the left and the score direction, Below-mean Quality, and missing-data symbols on the right. Both halves use matching rows with aligned labels and descriptions. Both groups keep fixed widths fitted to the minimum window width, and the guide's explanations remain on single lines as the window is resized. Below the table, Selected image shows a subframe's keep or reject decision, REVIEW status, Score, and main issue. For the master, it identifies its exclusion from scoring and ranking. Its compact measurement table aligns FWHM, eccentricity, and star count across the field, center, and four corners.

The bottom message area includes the visible and total subframe counts, the number of REVIEW candidates, and whether the master is included or hidden by the filter. These messages update when the selected group or REVIEW filter changes.

Click the Messages bar to collapse or expand the bottom message area on any tab. Status, review counts, score information and pixel scale continue updating while hidden. The window keeps its size, and the preference is saved with process-instance settings. Messages opens expanded by default. An invalid manually entered SQM time expands it to show the correction needed.

Click the Review information bar to collapse or expand Reading the table and Selected image together. Sort and filter remains available. Collapsing the information panels reduces scrolling or gives the review table more room without resizing the window or changing the selected image. This preference is saved with process-instance settings.

Hover over the selected-image assessment for the full explanation, or over its measurements for unavailable-region details. The filename is shown separately; hover over it for the full original or moved-file path and the registered-image path. Selecting another image updates these details without changing the window size.

Table 5

Column

Meaning

R1

Rank using the manual weights for field quality and its Below-mean Quality contribution, without corners. Rank 1 is worst.

R2

Rank using the manual weights for field and corner quality, including the corresponding Below-mean Quality contribution. Requires complete regional measurements. Rank 1 is worst.

R3

Rank using equally weighted, strongly correlated indicators other than PSF signal weight. Available when field measurements exist and at least one KPI qualifies, independently of the selected scoring mode. Rank 1 is worst.

R4

Rank of ranks, based on the equal-weight average of the normalized available R1, R2, and R3 ranks. R2 is omitted while corner scoring is off. Rank 1 is worst.

Reject

Your manual rejection choice. Clicking this column toggles the choice for an eligible original.

Review

REVIEW identifies a shortlist candidate or an image requiring inspection because of missing data. Hover over REVIEW to see the reasons for that image.

Image

The measured registered image or integrated master.

Acquisition time (UTC)

The complete date and time used for chronological sorting and quality plots. Master and unavailable timestamps are shown as a dash.

Score

The current quality score, from 0.000 to 1.000; higher is better.

PSF weight, FWHM (px), Stars

Field measurements used in scoring.

Sky brightness

Lowest mean of the center and four corner previews when corner scoring is enabled; otherwise the whole-image mean. Image units; lower is better.

Main issue

The enabled component with the largest weighted impact, such as Weak signal, Few stars, Large stars, Elongated stars, Large corner stars, High sky brightness, Medium guiding, Bad guiding, or Below-mean Quality. Good guiding is reported when no enabled quality penalty applies.

Field eccentricity

Field star elongation, from zero for round stars toward one for elongated stars. Lower is rounder.

Corner quality

The largest corner-quality penalty across the four corners. Lower is better. Stored results remain visible when corner scoring is turned off.

Below-mean Quality

Percentage of assessed indicators with worse-than-mean quality; lower is better.

Guiding quality

Combined guiding error: (minimum + 2 times average + maximum) / 4, in arcseconds. Lower is better. A value requires usable guiding coverage.

Guide avg, Guide min, Guide max

Arithmetic average, minimum, and maximum combined RA/Dec displacement during the exposure, in arcseconds. These are not RMS values. Hover for coverage and guiding quality. An asterisk marks partial statistics excluded from Score.

Worst corner

The corner contributing the largest regional penalty when corner scoring is enabled.

Original association

Whether the corresponding uncalibrated original has been verified for moving.

All manual Score components have columns: PSF weight, FWHM, Stars, Field eccentricity, Corner quality, Below-mean Quality, Guiding quality, and the selected Sky KPI. Sky brightness, Transparency and Sky quality all remain visible; their tooltips identify which one is selected for scoring. In correlation mode, columns for qualifying individual corner FWHM and eccentricity indicators appear at the right of the table; scroll horizontally to view them. Their tooltips report correlation and sample count. Missing measurements remain marked as unavailable. The quality-column tooltips explain their meaning and current inclusion in Score.

Main issue identifies a scoring contribution, not a diagnosis. For example, elongated stars may be associated with guiding, wind, or optical effects; inspect the image to decide whether to reject it. Hover over the score, issue, or quality columns for contribution details and missing-data information.

R1, R2, R3 and R4 rank subframes only. Their values remain group ranks when the table is filtered; they are not renumbered as visible row positions. Equal scores are resolved consistently by filename. The unscored master stays last when sorting by Score or any rank, in either direction.

Use Sort by in the Sort and filter area. Score, field-width, corner-quality, sky-background, and Below-mean Quality sorts show the poorest values first; PSF signal weight and star count show the smallest values first. R1, R2, R3 and R4 show rank 1 first. The button beside the selector shows Ascending or Descending. Click it to toggle the direction. Choosing another criterion restores that criterion's default direction. The selected criterion and direction are saved in process instances. Below-mean Quality keeps complete comparisons first, with Score resolving percentage ties in the chosen direction; incomplete comparisons follow in both directions.

Acquisition time starts with the earliest full date and time. Toggle to Descending for latest first. In both directions, dated subframes come first, undated subframes follow, and the master is last. Filename sequence numbers do not participate in this ordering. Sorting preserves the selected image and manual rejection choices. The quality plots retain their own order.

Choose Guide minimum, Guide maximum, or Guide average to sort by the corresponding guiding column. These start with the largest error first; use the direction button to put the smallest error first. Values are sorted numerically in arcseconds, including partial measurements marked with an asterisk. Unavailable values follow measured values in both directions, and the master is last. The choice and direction are saved in process instances and also determine the sequence for Rotate subframes (Review order).

Enable REVIEW candidates only to show just flagged subframes. This filter also hides the master. It changes table visibility without marking files for rejection or limiting the quality plots to those rows.

Hover over a REVIEW marker to see why the image was selected. For the Review worst percentage, the tooltip gives its group position, the rounded-up shortlist size, Score, and largest score penalties. It also lists any missing or insufficient measurements and guiding-data issues that require inspection. Both reasons appear when both apply. The explanation follows the current scoring settings and stays consistent when the table is sorted or filtered. REVIEW leaves the rejection decision to you.

When corner measurements are enabled, the REVIEW tooltip also explains unavailable preview measurements individually. It reports the detected-star count and required minimum, a missing or invalid FWHM or eccentricity result, or the recorded processing error. A preview with no result is described as not measured. When the measurement process supplies no detailed cause, the tooltip says so. These details also remain visible for missing regional indicators excluded by correlation scoring; their presence alone does not add a REVIEW flag.

Click a row to select it. Clicking the same selected row again deselects it. Keyboard navigation selects rows; Space toggles rejection of the selected eligible original. Double-click a row outside the Reject column to open it for blinking. Selection, REVIEW status, and rejection are separate states.

17 Quality indicators and acquisition order

[hide]

The Quality indicators tab shows one selected indicator at a time, with Score selected initially. The plots follow WBPP's per-frame style: each image occupies one bar position. They do not aggregate images into frequency bins.

The note beside the selected indicator shows whether higher or lower values are preferable and whether it is included in Score under the current settings. A zero weight, disabled corner scoring, or excluded guiding is identified. In correlation mode, each corner FWHM and eccentricity reports its own inclusion or exclusion. In manual mode these measurements contribute through Corner quality; Center FWHM is used in corner/center ratios. Guide minimum, average, and maximum contribute through combined guiding quality when coverage is usable. Score itself is the combined result. Plotted measurements remain available even when their scoring contribution is disabled.

The blue information symbol button opens the selected histogram’s explanation, including its definition and relevant measurement or calibration details. Reading the plot uses labeled rows for inspecting bars, ordering and the vertical axis. A color-and-marker legend separates image bars, reference lines and quality shading, with a sample and written meaning for each. SQM and autofocus markers appear only when relevant to that plot. These explanations are kept in the information window, without repeating them in plot headings or tooltips. The direction of quality and current Score inclusion remain visible beside the selector. The footer shows only the frame count and order. Hover over individual bars for their image-specific values.

Choose Order by: KPI value for ascending values, or Acquisition time for the acquisition sequence. The plot order is independent of the review table. In chronological plots the master is first, dated subframes run from earliest to latest, and undated subframes follow. Bar spacing is uniform; the horizontal axis is an image sequence, not a proportional elapsed-time axis.

Table 7

Color or marker

Meaning

Cyan

Currently selected image. This highlight takes precedence over the normal bar color.

Purple

REVIEW candidate.

Blue

Other subframe.

Red

Manually rejected or moved original.

Gold

Master, with a master marker.

Green dashed line

Mean of available values for subframes that have not been rejected or moved. The master is excluded.

Orange dashed line and AF tag

Autofocus start within the current group's acquisition-time range, with the final focus position when available.

Hover over a bar to see its filename, value, status, and acquisition time. Click it to select the corresponding row in Review and measurements. If necessary, the script clears REVIEW candidates only to reveal the selected image, including the master. Clicking a plot bar does not change rejection decisions.

Except for the category plots described below, the vertical range follows the subframe values with approximately 10% padding. An outlying master value is clipped at the plot boundary and marked with an arrow; its actual value remains visible. This keeps subframe differences readable when the integrated master has a much larger PSF signal weight. Missing values retain their place in the sequence but have no value bar.

Field, center, and corner FWHM plots can show Good, Medium, and Bad background bands using your telescope and reference observing conditions. The FWHM grading settings icon is shown only for Field, Center and corner FWHM plots. Open it to configure the limits. The FWHM grading section below explains the parameters and calculation.

Field, center, and corner eccentricity plots follow the plotted values, including the master, with about 10% headroom and a small lower margin, bounded by zero and one. Green, amber, and red background bands are clipped to the visible range. Good eccentricity is at or below 0.42; Medium is above 0.42 and at or below 0.60; Bad is above 0.60. The legend and dotted lines mark these boundaries in both KPI-value and acquisition-time order. Hover over a bar to see its category. Eccentricity is dimensionless: zero is round and values approaching one indicate increasingly elongated stars. These ranges are practical inspection guidelines. Score continues to compare eccentricity within the group; the bands do not change scoring or automatically reject an image.

For Guide minimum, Guide average, and Guide maximum, green, amber, and red background bands mark Good, Medium, and Bad guiding. The legend gives the boundaries in arcseconds, and dotted lines separate the bands. Guide minimum and Guide average follow their measured values with a small margin above and below, like Guiding quality. Only the portions of the bands within that range are shaded. Guide maximum includes zero and all three categories, as well as the measured values. These bands appear in both KPI-value and acquisition-time order when the imaging scale is available. The bar colors still identify selection, review, and rejection status. No category boundaries are drawn when the imaging scale is unavailable.

The Guiding quality plot shows the combined indicator without category bands. Its information window gives the definition: G = (minimum + 2 times average + maximum) / 4, in arcseconds. The average contributes 50 percent, and the minimum and maximum each contribute 25 percent. Lower values are better. The displayed minimum, average, and maximum categories help inspect individual guiding measurements; the combined value G determines the guiding penalty used in Score.

The Score and Below-mean Quality histograms show subframes only; the master is omitted.

The green plot mean excludes rejected subframes and can change as you make decisions. The scoring baselines and the arithmetic means behind Below-mean Quality use all supported subframes in the group, including rejects. These intentionally serve different purposes.

Acquisition timestamps come first from the verified original's DATE-OBS or DATEOBS metadata, then from the registered image. Separate TIME-OBS or TIMEOBS fields are supported when the date field contains only a date. Explicit timezone offsets are converted to UTC; timestamps without an offset are interpreted as UTC. The table shows the same UTC timestamp used by the plots.

Acquisition filenames may show a local clock or an end time while the header records a UTC exposure start. The displayed Acquisition time (UTC) is the authoritative value used by this script. The filename counter suffix is ignored.

18 Sky-background comparison

[hide]

With Session SQM supplied, enter the reading's local date and time beside it. The script automatically estimates the Bortle class and selects the eligible subframe whose exposure midpoint is nearest to that time in each directly calibrated group. Masters, rejected or moved frames, and frames without usable exposure times or backgrounds are excluded. Equal time differences prefer the earlier exposure. Click the information symbol button on the Quality indicators tab to open the reference frame, reading time, exposure midpoint in UTC, time difference and reference background. With Gaia calibration, this information window shows the reference, accepted-star count and calibration scatter. It also lists the SQM-to-Bortle ranges used by the script, including exact boundary handling and the shared range for classes 8–9. The table adds the calibrated sky-background range in the plot's image units and the shading color for each class. The shading legend identifies the Bortle classes covered by each color. Before calibration, the SQM ranges remain available; calibrated values and shading show a dash.

The selected frame's measured background is assigned the SQM reading directly. Other backgrounds are converted relative to it. Gaia is not used when an SQM reading is supplied. Missing or ambiguous reading times leave calibration unavailable; enter the complete date/time and check the UTC offset. A missing SQM time never switches to Gaia automatically.

Separate red and blue filter groups also show sky-background shading and the SQM line. They use the green group's calibration, adjusted to their own background levels using the closest-in-time pair of eligible green and red or blue exposures. This uses existing measurements; no additional photometry is needed. The plot caption says relative to green, and the information button shows the calibration reference, paired exposures, time gap and measured backgrounds. The calibrated Bortle table uses the selected filter's image units.

This transfer assumes similar sky conditions between the paired exposures and a stable relative filter response across the group. It is a relative estimate, not independent red/blue sky photometry. Large time gaps, changing sky colour, clouds or transparency can affect it. Red and blue shading requires a compatible measured green group with usable timestamps. Without one, the information button explains why shading is unavailable. With no SQM reading, red and blue reuse the green group's Gaia calibration.

The compact plot caption shows the Bortle class and SQM reading or Gaia calibration. The footer shows the frame count and plot order. The vertical axis follows the plotted backgrounds, including the master when present, with a small margin below and headroom above. Bortle limits do not extend it. Only the parts of the quality bands inside that range are shaded.

The pink dashed line marks the SQM reading on the image-unit axis. Bars below it indicate darker backgrounds; bars above it indicate brighter backgrounds. With corner scoring enabled, calibration uses the same minimum preview mean as the plot; otherwise it uses the whole-image mean. Switching this basis reuses the measured backgrounds.

Without an SQM reading, choose Sky brightness on the Quality indicators tab. The Bortle class selector appears immediately to the left of Calibrate sky. Local Gaia photometry calibrates the unrejected subframe with the best default field-only score in each directly calibrated group. Changing score weights or choosing R2, R3 or R4 does not change that selection rule. Choose — to disable Bortle shading and automatic Gaia calibration.

Calibration runs after image measurements or when the required sky settings are supplied afterwards. Existing field and corner measurements are retained. Calibrate sky is shown only while the current group has no usable calibration; it disappears after success and reappears if calibration becomes unavailable. Changing Bortle class reuses Gaia calibration. Changing the SQM reading or time updates the direct calibration using existing background measurements. Returning to a previously calibrated mode reuses its result when still applicable.

The plot is shaded automatically when calibration and a Bortle class are available:

Table 8

Category

Meaning

Good, green

Darker background corresponding to lower Bortle classes.

Medium, amber

Background within the selected or inferred class's approximate brightness range.

Bad, red

Brighter background corresponding to higher Bortle classes.

For example, class 6 uses approximately 18.94–19.50 mag/arcsec² for Medium. Class 1 has no lower-class Good band. Classes 8 and 9 share a brightness range below 18.38 mag/arcsec². An SQM reading in that range is reported as Bortle 8–9; it cannot distinguish the two visual classes. The manual selector retains separate choices for 8 and 9.

Hover a subframe bar to see its estimated brightness and category. The integrated master is a comparison image and receives no brightness estimate. The SQM line and Bortle shading do not change Score, REVIEW or rejection decisions.

Use linear, calibrated WBPP registered frames. Luminance, green and V filters support direct calibration; RGB uses the green channel. Separate red and blue groups use green as described above. A shared calibration within each group requires matching exposure, camera, gain, binning, filter, pedestal and image dimensions. The green-to-colour transfer requires the same camera, gain, binning, image dimensions and target metadata. Other filters remain uncalibrated. Without SQM, the direct reference also needs sky-coordinate information and a local Gaia DR3/SP, DR3 or EDR3 catalog configured in PixInsight. Unreliable coordinates are rebuilt on a temporary image if necessary; source files stay unchanged.

The SQM reading should represent the sky in the target's direction near the matched exposure. An SQM-anchored plot is a relative estimate, not an independent check of the meter reading. Stars, nebulosity, gradients and changing transparency can affect the measured image background. Gaia gives a V-equivalent estimate, which differs from an SQM meter's passband. Bortle ranges are approximate.

Process icons save the reading, its local date/time and the manual Bortle choice. They do not save measurements or calibration; closing the script discards the measured session.

See the approximate Bortle comparison table and Gaia photometric transformations.

19 FWHM grading

[hide]

Use FWHM grading on the Quality indicators tab to set personalized Good, Medium, and Bad limits. Enter the telescope aperture in Input files and observing conditions; the FWHM dialog shows this value for reference. The script uses its validated image pixel scale to convert the calculated angular limits to pixels, matching the FWHM measurements. Bands are unavailable until both aperture and pixel scale are available. The X and Y pixel scales must agree within 0.1%. Clear Show FWHM quality bands to hide the shading. The dialog uses a single column of aligned controls: Telescope and image summarizes the current setup, Observing conditions holds altitude, wavelength and optional optical blur, and the next section sets the Good and Medium boundaries. The result shows three distinct Good, Medium and Bad rows in pixels. The information icon opens the model details, organized into labeled tables for current grading limits, reference conditions, blur contributions and filter defaults.

Table 9

Parameter

Initial value

Meaning

Aperture

Unset

Telescope aperture diameter in millimeters, entered in Input files and observing conditions.

Zenith seeing

0 arcsec

Optional estimate entered beside Aperture; zero selects fixed seeing limits.

Good up to, relative mode

115%

Upper Good boundary as a percentage of expected FWHM.

Medium up to, relative mode

135%

Upper Medium boundary as a percentage of expected FWHM; must exceed the Good percentage.

Good up to, fixed mode

1.5 arcsec

Upper seeing limit for Good quality, specified at the zenith and 550 nm.

Medium up to, fixed mode

2.5 arcsec

Upper seeing limit for Medium quality at the same reference conditions; must exceed the Good limit.

Target altitude

45 degrees

Fixed target altitude used for the grading model; allowed range 15–90 degrees.

Wavelength

From filter

Editable wavelength for the selected filter; unknown filters start at 550 nm.

Additional optical blur

0 arcsec

Optional additional blur from residual focus errors or optical aberrations; zero adds none.

With a positive Zenith seeing estimate, the FWHM grading dialog shows adjustable percentage limits. The initial Good and Medium limits are 115% and 135% of expected FWHM. These are chosen acceptance margins, not physical standards. Good means close to the expected performance under the supplied conditions; it does not mean equally sharp images under different seeing conditions.

With Zenith seeing set to zero, the dialog shows the two fixed seeing limits instead. These are adjustable acceptance boundaries, not measurements of the current atmosphere. The settings for both modes are retained when switching between them. Choose a target altitude appropriate for the session. A larger optical allowance permits broader stars. Guide-camera data are excluded from the FWHM grading model.

Effective wavelength starts from the selected filter. You can edit it to suit your filter; the chosen value is remembered for that filter and saved with the process settings. Switching to another filter uses that filter's saved value or its default. The Use filter default icon restores the value derived from the filter name. Names are matched without regard to capitalization:

Table 10

Filter names

Wavelength

L, Luminance, G, Green

550 nm

R, Red

650 nm

B, Blue

450 nm

H, Ha, Halpha, H-alpha

656.3 nm

O, OIII, O3

500.7 nm

S, SII, S2

672.4 nm

Unknown or unavailable filter

550 nm

Broadband values represent typical wavelengths; the exact effective wavelength also depends on the filter, camera, and source spectrum. The SII value represents the doublet near 671.7 and 673.0 nm. See SII wavelengths.

Additional optical blur represents additional blur from the imaging optics beyond the diffraction and seeing terms. Leave it at zero unless you have an independent estimate. ASIAIR autofocus star sizes are HFD values in pixels and include several sources of blur; they cannot reliably determine this extra optical term. Star sizes from a guide camera also do not describe the imaging system's optical blur. See ZWO's explanation of autofocus star-size units.

The expected blur is calculated from the telescope, atmosphere, and optional optical allowance:

Total FWHM = square root of (Diffraction² + Corrected seeing² + Additional optical blur²).

Diffraction = 1.028 × 206264.806 × wavelength in nm × 0.000001 / aperture in mm.

Corrected seeing = zenith seeing × sin(target altitude)^(−0.6) × (wavelength / 550)^(−0.2).

In relative mode, use the supplied Zenith seeing to calculate expected FWHM. Good limit = expected FWHM × Good percentage / 100; Medium limit = expected FWHM × Medium percentage / 100. In fixed mode, calculate one total FWHM from each of the Good and Medium seeing limits.

The model calculates angular limits internally. Each displayed limit is converted to pixels by dividing by the script's image scale in arcseconds per pixel. The settings preview, plot labels, and measured FWHM use the same pixel units.

Table 11

Category

Measured FWHM

Good, green

At or below the calculated Good limit.

Medium, amber

Above the Good limit and at or below the Medium limit.

Bad, red

Above the Medium limit.

The bands apply to Field FWHM, Center FWHM, and the four corner FWHM plots in either plot order. FWHM measurements, quality limits and axis ticks use one decimal place. The chart header shows the indicator name and category limits. Click the information symbol button on the Quality indicators tab for the expected FWHM and observing-condition details. The button appears only when information is available for the selected indicator. Measurements and category comparisons retain their full precision; displayed rounding does not change scores. Changing the parameters refreshes the plots without repeating measurements, and the parameters are saved with process-instance settings. These bands are visual inspection guides: they do not change Score, REVIEW flags, or rejection decisions.

The root-sum-square model is exact for independent Gaussian blur components. Actual Airy and Moffat star profiles, optical aberrations, image resampling, and undersampling can depart from this approximation. Check the limits against representative images from your setup and adjust the acceptance limits as needed.

Background: ESO image-quality model and GTC seeing corrections. This script uses 550 nm as the reference wavelength for its zenith-seeing inputs.

20 Correlation plot

[hide]

Open the Correlation plot tab and choose any two quality indicators with X axis and Y axis. The plot and Pearson correlation coefficient update immediately. Swap axes exchanges the selections without changing the correlation. The initial pair is Field FWHM on X and PSF signal weight on Y.

Regional KPIs appear in both selectors only when corresponding subframe measurements are available in the current group. Click Measure corners to measure missing regions and add these choices. If a selected KPI becomes unavailable after changing groups or clearing measurements, that axis returns to its initial selection; the other axis is retained when available.

Each point represents one subframe with both measurements available. The plot and correlation use the same subframes in the selected filter group, including rejected frames, independently of the review table's filter and sort order. The master is always excluded. Frames missing either measurement are omitted, and the count below the plot reports how many could not be plotted. Click Measure corners for regional indicators; load matching guiding logs for guiding indicators. Guiding values use arcseconds, FWHM uses pixels, and Below-mean Quality uses percentages.

Hover over a point to see its filename and both values. Click it to select the image in Review and measurements. If necessary, the REVIEW-only filter is cleared to reveal that image. Cyan identifies the selected image, purple a REVIEW candidate, blue another kept frame, and red a rejected frame. Plot selection does not change rejection decisions. Identical measurements can overlap; keyboard focus and Enter also select points.

The highlighted banner above the plot displays Correlation (Pearson) to four decimal places and its interpretation. Hover over the banner to see the full precision value. Strength and direction use the unrounded coefficient. The coefficient r describes the strength and direction of the linear relationship between the two KPI values. Positive means both increase together; negative means Y tends to decrease as X increases. Direction refers to the values, not whether image quality improves. The display gives r and its meaning using these continuous bands:

Table 12

Absolute r value

Strength

1

Perfect

0.7 up to, but not including, 1

Strong

0.3 up to, but not including, 0.7

Moderate

Above 0 and below 0.3

Weak

0

None: no linear relationship

The sign of r determines the positive or negative direction. Zero means no linear relationship; a nonlinear relationship may still exist. With fewer than two paired subframes, or no variation in either selected KPI, r is unavailable and the display explains why.

Axis choices are saved in process instances. Changing the plot uses the existing measurements and does not require another measurement run.

21 Log folder, timing, and autofocus markers

[hide]

Select Log folder in the directory section. All Autorun_log and PHD2_GuideLog files are read directly from this folder, regardless of letter case. Include continuation logs if acquisition was restarted. Files marked _CHN are ignored because they are translated copies. Subfolders and unrelated logs are not processed. Select the folder again to refresh its contents.

Autofocus starts from all matching logs are combined in chronological order. Overlapping logs produce one marker per start. An unreadable log is listed in Autofocus times; readable logs are still used. The report lists the source filenames for every start. If duplicate logs report different final positions for the same start, the position is marked as conflicting and the plot shows AF without a position.

The script reads lines such as 2026/09/05 23:08:11 [AutoFocus|Begin] followed by the autofocus trigger description. It pairs each start with the final successful focus position from that run. For example, the later line 2026/09/05 23:11:19 Auto focus succeeded, the focused position is 692 produces the label AF 692 at the 23:08:11 start. Intermediate EAF positions are ignored. If no successful position is reported, the label remains AF. Repeated entries with the same start are deduplicated, and invalid start timestamps are reported as skipped.

Both log types use local time. Infer from image timestamps compares the local timestamp in each subframe filename with its UTC exposure-start metadata. It accounts for filenames that record exposure completion. A single unambiguous inferred offset fills the UTC offset field as its default. Select Custom UTC offset to edit this value. In automatic mode the field remains visible but is read-only. The filename timestamp must have the form YYYYMMDD-HHMMSS or YYYYMMDD_HHMMSS.

If no single offset is available, the field shows a dash. Its tooltip explains missing or multiple offsets. When several offsets are known, automatic mode continues to apply them to matching exposures. For a manual setting, choose Custom UTC offset and enter local time minus UTC in hours, such as +2 for UTC+02:00 or +5.5 for UTC+05:30. The current computer's timezone is never assumed. At a daylight-saving transition, ambiguous times remain unresolved rather than being assigned to an exposure by guesswork.

The console reports all extracted autofocus starts as written in the log and converted to UTC, together with each final focus position or not reported. Click Autofocus times to view and copy the same list. Changing the folder or timestamp convention updates the report and markers without remeasuring the images. Clearing the folder removes the markers and retains image measurements.

Autofocus markers appear only when image-quality plots use Acquisition time order. The guiding minimum, maximum, average, and quality plots omit autofocus markers and their legend. Only starts within the earliest-to-latest known subframe timestamps of the current group are shown. A start between two subframes is placed between their bar positions. Hover over the AF tag to read its log time, UTC time, final focus position, and trigger description. Markers use the start time even though the final position is reported later; no end time or duration is displayed.

All extracted starts remain in the report, including starts outside the current group's time range. If markers appear at unexpected positions, compare the UTC acquisition column with the converted times in Autofocus times and check the log's timezone setting.

22 Guiding measurements and quality

[hide]

The script matches PHD2 samples to each subframe's UTC exposure start and duration. Guide min, Guide max, and Guide avg report the combined RA/Dec displacement in arcseconds. The guider's recorded pixel scale converts its measurements to arcseconds; the imaging camera's pixel scale sets the quality thresholds.

Let p be the image pixel scale in arcseconds per pixel. The guiding quality value is:

G = (minimum + 2 * average + maximum) / 4

The average contributes 50 percent, and the minimum and maximum each contribute 25 percent. This balances typical performance with the observed extremes. For unequal X/Y image scales, the smaller available scale is used. The UI displays the Good, Medium, and Bad boundaries in arcseconds for your setup. For example, an image scale of approximately 0.507 arcseconds per pixel gives a Good limit of approximately 0.507 arcseconds and a Medium limit of approximately 0.887 arcseconds; larger errors are Bad.

Table 13

Category

Guiding quality value

Good guiding

G is equal to or below p.

Medium guiding

G is above p and equal to or below 1.75 times p.

Bad guiding

G is above 1.75 times p.

Guiding adds a penalty of (G / p minus 1) divided by 0.75, bounded between zero and 10. Good guiding therefore has zero penalty; the upper medium boundary has a penalty of one. Weight in the Guiding group sets its relative Score weight, with a default of 25. Set it to zero to exclude guiding from Score while retaining the measurements. Guiding statistics, combined guiding quality G and their displayed thresholds are rounded to one decimal place in arcseconds. The review table, plots and guiding tooltips use the same precision. Calculations, sorting and category comparisons use the full-precision values.

Choose how Guiding in Score handles the measurements:

Table 14

Mode

Effect

Require guiding data

Use valid guiding measurements in Score. Missing or incomplete guiding data flags the subframe for review.

Use available guiding data

Default. Use valid guiding measurements. For a subframe without a usable guiding KPI, omit guiding from its Score and Below-mean Quality calculation. Missing guiding alone does not flag it for review.

Ignore guiding KPI

Exclude guiding from every Score, Below-mean Quality calculation, and review flag. The measurements remain visible for inspection.

Changing this mode recalculates the rankings immediately, preserves the measurements and manual rejection choices, and is saved in process instances. Image-quality issues can still flag frames for review in every mode.

Only valid guide measurements are used. Lost-star and rejected samples do not count as zero error. Settling samples inside an actual exposure remain part of its measurements. Guiding needs at least three valid samples, at least 90 percent time coverage, and no long uncovered interval or conflicting duplicate samples to contribute to Score. Partial statistics are marked with an asterisk. Missing timestamps, an unknown image pixel scale, or insufficient coverage require inspection in Require guiding data mode. Rejected guide samples also flag an exposure for review when its guiding KPI is in use.

Hover over a guiding column for its category, sample count, coverage, and thresholds. The selected-row details also show the guiding category or the reason it is unavailable. If an exposure is outside the loaded log coverage, add the missing continuation log and reselect the folder. Refreshing logs or changing their offset updates guiding and rankings without repeating image measurements or changing manual rejection choices.

23 Move rejected originals and rerun WBPP

[hide]

The green checkmark opens a move-plan review. It includes only manually rejected subframes with verified original files. Review the listed filenames and destination before confirming.

The script checks the complete plan before moving files. It verifies the original associations and file fingerprints, preserves each original's relative directory structure, and refuses to overwrite destination files. The calibrated images, registered images, masters, and calibration originals are not moved.

A move journal named SQR-moves followed by a timestamp and a .json extension is saved in the rejection destination. Keep it with the moved files so that you can undo the move later.

To restore a moved original, select its row and click the Undo arrow beside the green checkmark. The Undo moved originals dialog preselects that file and shows its original path. Select additional files if needed, then click Restore selected. Restored files return to their original folders and their rejection marks are cleared in the current review.

Undo also works after rescanning or reopening the script. Scan files or import WBPP, and specify the Rejected originals folder to enable the Undo arrow, then choose the original SQR-moves journal from the rejection folder and select the files to restore. Select available checks all eligible files; Clear selection lets you start with none. You do not need to measure the images again to restore them.

Files that have changed, are missing, or have an occupied original path cannot be restored. Existing files are never overwritten. Already restored files are identified in the list. If restoration stops partway through, reopen the same move journal to restore the remaining eligible files. Each restoration also saves an SQR-undo journal in the rejection folder.

After moving, rerun WBPP from the remaining originals. With an imported setup, use Create a reviewed WBPP instance to update its light list. Otherwise refresh the inputs manually and exclude the rejection directory. Use an output directory appropriate for the new run so that you can distinguish the new integration from the reviewed run.

24 Common questions

[hide]

Table 15

Situation

What to check

The master is missing

Check for an uncropped masterLight image in the WBPP master directory, with a filter identifier matching the group. Multiple possible masters require an unambiguous group-name match.

An original cannot be rejected or moved

Read Original association and the row details. Check that the original-light root contains the correct exposure and that matching metadata has not been removed or changed.

There are more REVIEW rows than the chosen percentage

Insufficient measurements also require review. Check sparse regions, Minimum stars, and missing eccentricity or corner data.

R2 or corner plots are unavailable

Click Measure corners beside the magnifier. R2 requires complete regional measurements; a regional plot requires a valid value for that indicator.

A score looks good despite missing data

Inspect its missing-data explanation. A partial score is not a complete quality assessment; missing required data is flagged for review.

The master bar is clipped

This is expected when it lies outside the subframe plotting range. Read the displayed master value or hover over its bar.

Acquisition times differ from filenames

Check whether the filename uses local time or exposure end time. The script displays and orders the full metadata timestamp in UTC.

No autofocus markers appear

Guiding plots intentionally omit autofocus information. For image-quality plots, use Acquisition time order, verify the log timezone, and check that extracted starts fall within the group's known subframe time range.

Closing and reopening loses measurements

Review data and decisions are held only in the open dialog. A saved process instance restores settings, not measured rows.

25 File locations and network access

[hide]

File saves, temporary writes, and image move/undo operations reject locations inside the PixInsight installation. On Windows, choose direct local drive paths; network shares, substituted drive letters, symbolic links, junctions, and ambiguous device names are refused. Ordinary cloud placeholders are supported. On Linux and macOS, resolved links and directory identities are checked against the installation.

On Windows, file protection runs the operating system utility mountvol.exe from the Windows System32 directory, using the read-only /L option to identify volumes and reject alternate drive mappings. It does not change mounts, request administrator privileges, process images, or use a command shell.

Replacing a text or data file briefly creates verified .tgscripts-write- and .tgscripts-recovery- files with unique names beside that file. The outside directory entry is replaced without writing through a pre-existing hard link. Successful operations remove these temporary files. If restoration fails, the error identifies the retained recovery file; keep it until the previous contents have been recovered.

Corner measurements write temporary XISF files in uniquely named .sqr-work directories under the operating system temporary directory. Rejection recommendation trials use .sqr-recommendation directories there. The script removes this scratch data when these operations finish.

Reviewed WBPP setups are saved at the .xpsm paths you choose. Confirmed moves place rejected original images in the selected rejection folder and write SQR-moves JSON journals there. Undo restores originals to their recorded source paths and writes SQR-undo JSON journals in the rejection folder.

Analysis uses local images, acquisition logs, and configured local Gaia data. The script makes no network connections. The external executable launch used to load an exported WBPP icon is described in Create a reviewed WBPP instance.