Apply a Common Crop to Aligned Masters with DynamicCrop
Remove stacking borders while keeping monochrome channels on the same pixel grid. Choose a crop inside the shared usable field, save one DynamicCrop instance, and verify every output before gradient correction and combination.
Sources reviewed October 25, 2026
Why the crop needs to be shared
Registration and stacking can leave dark borders, incomplete coverage, or noisy edge strips. A preparation crop removes these before gradient modeling and channel combination. It is different from resizing a finished image for the web.
Aligned R/G/B, S II/Hα/O III, and optional luminance masters must retain the same sky coordinates at corresponding pixel positions. Drawing a separate crop by eye on each channel can shift those positions even if the outputs have identical dimensions.
This exercise uses one saved, unrotated, unscaled DynamicCrop instance on inputs that already share dimensions and a registered pixel grid. Cropping does not align data. If the starting geometry differs, register to a common reference and verify that result first.
1. Preserve the full integrations
Save the original WBPP integrations and make clearly named working copies, for example Target_R_cropwork, Target_G_cropwork, and Target_B_cropwork. For SHO use the three actual emission-line masters; for HOO use Hα and O III once each. Include L if you intend a later LRGB combination.
Keep these masters linear and stars present. Use STF for display without applying a permanent stretch. Check their dimensions, orientation, and star alignment at 1:1 in the center and corners. Equal width and height alone do not establish registration.
For one-shot-color broadband data, the RGB integration is a single crop target. If combining it later with separate luminance, narrowband data, or RGB stars, include every relevant aligned branch in the common-geometry plan.
2. Find the shared usable field
Inspect all intended masters with Blink or matching views. Use a display stretch strong enough to reveal edge defects, then check at 1:1. Dithering, rotation, or different session coverage can leave a different boundary in each filter.
Choose a rectangle inside the intersection of all usable areas. Let the most restrictive channel define each edge. A border need not be fully black to be unsuitable: inspect abrupt noise changes or low-coverage strips as well.
Retain as much useful field as practical, but do not preserve a damaged edge just to keep the original dimensions. If removing it would sacrifice an important part of the subject, review registration, selected exposures, or integration coverage before accepting the compromise.
3. Define one crop without resizing
Open DynamicCrop through All Processes and reset a reused instance. Use one working master to define the shared rectangle, checking it against the boundaries you found in the other channels.
- Drag a rectangle over the active master to mark the region to keep. Adjust the edges and inspect the Size/Position values. The rectangle must remain inside valid coverage in every channel.
- Keep Rotation at 0 degrees and Scale X and Scale Y at 1.0000 for this preparation exercise. Check the expected output pixel dimensions. Avoid changing the Scale output-size fields to a web or print size.
- Check the final rectangle at full view and at each edge. Do not extend it beyond the source boundary; filled pixels are not additional sky coverage.
- Before executing, drag the New Instance triangle onto the workspace to save the configured crop. Give the instance a clear name such as Target_common_crop and save the process icon file for later reuse.
- Record the original input dimensions and the crop’s position, size, rotation, and scale. The saved instance is intended for these matching original inputs, not for arbitrary images or already cropped outputs.
4. Apply that exact instance once per master
Execute the defined crop on the first working master using DynamicCrop’s execution control. End its active dynamic session before applying the saved instance elsewhere. Retain the saved icon without redefining the rectangle on a changed image.
Drag the saved common-crop instance onto each remaining original-size working master. Apply it once to each. Keep a checklist of processed views so a cropped output does not accidentally receive the same operation again.
Do not redraw the crop or reopen and change the saved instance between channels. DynamicCrop stores geometry relative to the target; matching starting dimensions and geometry are essential for the same settings to select the same pixels.
Save outputs separately as Target_R_cropped.xisf, Target_Ha_cropped.xisf, or equivalent filter-specific names. Keep uncropped originals so another field choice starts from a clean checkpoint.
5. Verify more than the dimensions
Confirm that every output has the same width and height. Then compare stars or compact structures at corresponding pixel positions with Blink or matching views. They should remain aligned through the center and corners, with no new offset or scale difference.
Inspect all four output edges for leftover invalid pixels, seams, or noisy coverage bands. If one channel still has an unsuitable border, redefine a common crop from the uncropped masters and repeat the entire set rather than trimming that channel alone.
For HOO, reuse the single cropped O III master in both green and blue when combining later. For LRGB, verify L against the RGB inputs too. A common filename suffix does not prove a common crop.
6. Continue to gradient correction and combination
Save the accepted cropped linear masters before further edits. Continue to the gradient stage, using the filter-appropriate MGC or DBE workflow. Cropping removes invalid edges; it does not remove genuine sky gradients, dust shadows, or reflections.
Check astrometric metadata after any geometry change. Do not assume a previous sky solution remains valid on the cropped image. For MGC, establish valid astrometry at the final geometry and run SPFC on the actual current master before correction.
Existing masks and saved DBE sample positions can also be tied to the old geometry. Rebuild or verify them before reuse. Do not independently alter the geometry of one channel during later processing.
After accepted gradient correction, optionally test LinearFit and combine RGB, SHO, or HOO using the linked lessons. Keep preparation cropping separate from a later artistic crop, and apply any later shared-branch geometry change consistently.
Quick troubleshooting
Outputs have equal sizes but stars do not line up: check the original registration, crop origin, rotation, and whether one input was already cropped. Resizing to equal dimensions does not repair registration.
The saved icon selects a different region: verify that all targets began with identical geometry. Recreate the crop on the intended uncropped grid rather than manually nudging each result to look close.
The image became smaller than expected or was resampled: inspect Scale X/Y, Rotation, and the input state. Start again from the original copy with scale 1 and zero rotation.
A black strip remains: inspect the shared coverage in every master and revise the common rectangle. Do not fill or clone the border to make gradient modeling accept it.
The sky solution fails afterward: solve the current cropped image with the correct effective image scale and verify the report before coordinate-dependent processing.
Keep the full integrations, cropped masters, saved crop instance, input dimensions, and verification notes. This is an original practice sequence; the linked worked examples explain crop geometry and common-channel processing, while installed help governs the current interface.