
RawTherapee can remove a narrow border during demosaicing before an export-size decision is made. Its Border control can be set to 0 when the full RAW frame must be retained, but the outer pixels may then show reconstruction artifacts. Increasing the export dimensions afterward does not restore the discarded sensor-edge coverage.
Establish whether the missing area is a processing border
Record the RAW frame's expected dimensions and the reason they must remain exact. A fixed-size DNG frame used in a video sequence has a different requirement from a still photograph that will be cropped for composition.
Check the processing setup for an intentional crop and output resizing as separate stages. This guide does not diagnose every smaller export as a Border setting. It isolates a documented RAW reconstruction behavior.
Pixel count also differs from physical print size. The pixels-to-print-size guide explains why changing print resolution metadata does not supply missing picture rows.
Decide whether full edge coverage is necessary
The RawPedia Border section explains that peripheral samples lack the surrounding neighborhood available to interior pixels. Its default crop avoids relying on those less favorable edge calculations.
Usually that default is appropriate. Override it only when the production requirement actually needs the original frame boundary. RawPedia gives fixed-size 1920×1080 RAW DNG frames as an example, rather than recommending zero for every photograph.
Write down the acceptance condition before changing the value: exact frame coverage, an approved pixel count and an acceptable outer edge. Those conditions must all be checked independently.
Compare the border override on a working copy
Locate Border in the demosaicing controls and retain the original processing state. Set it to 0 for the proposed comparison. Hold crop and output resizing steady so a later operation does not hide the difference.
Inspect the saved result's dimensions and its outermost rows and columns. Look at corners, high-contrast details touching an edge and any edge that must align with adjacent frames. A numerically correct canvas can still contain objectionable artifacts.
If the image is then fitted into an editor, distinguish that fitting from the RAW border decision. The Premiere frame-fitting guide describes downstream placement, which cannot establish what RawTherapee reconstructed.
Accept coverage without promising edge quality
For a sequence, inspect more than one frame before adopting the setting across the job. Detail and illumination near the edge can vary.
If zero preserves the required extent but the border is unusable, keep that tradeoff explicit. Do not silently enlarge or invent edge content and label it recovered RAW data. Reconsider the production framing requirement or obtain a suitable source. This guide proposes a comparison; it reports no measured DNG conversion or verified frame dimensions.
Related tasks
- How Large Can You Print a Generated Image?
- Why Does a Large Photo Lose Detail After I Scale It to the Premiere Frame and Zoom In?
Sources and scope
- RawPedia: Demosaicing — Border: Default demosaicing crops peripheral rows/columns because border samples lack surrounding pixels. Border 0 disables that crop but can introduce edge artifacts; the documentation gives fixed 1920x1080 raw DNG frames as a reason to require it.
Official RawPedia and Siril documentation was opened and read on October 23, 2026 (Asia/Shanghai). The Siril calibration reference identifies version 1.4.4. All acquisition, processing and acceptance examples are proposed; no camera, RAW file, calibration sequence or command was tested. The cover is decorative.
Prepared by Imagild Editorial, the company editorial team responsible for these guides. AI assisted research and drafting. This is documentation-based guidance with proposed examples, not a claim of hands-on testing or an independent product review. Report a correction.
Cover: Mariya Eskina / Pexels, Pexels License. This is an illustrative source photograph, not an application screenshot or a generated test result.