mirror of
https://github.com/storytold/pdfcraft.git
synced 2026-11-06 18:52:36 +00:00
huppifluppandClaude Opus 5.5
308d22cbad
Scan & OCR ▸ Enhance scanned file (Acrobat's Enhance ▸ Scanned document) now cleans up scanned pages: background removal (the camera clean-up's level 0-100), edge shadow removal (book gutters, copier borders) and despeckle, each its own switch. The camera tool renders the page and replaces it with a picture, which would drop a scan's invisible OCR text, comments and links. Instead each scan image (pictures covering at least half the page, each at least a tenth) is decoded, cleaned and written back into its own image object at the same size and colour space: a JPEG stays a JPEG (quality 90), other grey/RGB images are stored losslessly, a 1-bit image stays 1-bit with its decode array, and every other dictionary entry (SMask, Interpolate, ...) is kept. An image several pages share is cleaned once. What can't be cleaned (CMYK, indexed, JPEG 2000, JBIG2, CCITT, colour-key masks, broken data, oversized images) is reported, never skipped silently; pages that are no scan are listed too. The engine works as a job like Optimize (scan_clean_job, run on a worker with progress and cancel, apply_scan_clean refuses a result for an older document state) and applies one undoable step, "Enhance scanned file". pdfcraft-optimize exposes decode_samples/encode_samples for in-place pixel edits. camera_clean gains clean_samples for grey and RGB samples with separate switches and an edge-shadow pass; the camera tool's output is unchanged. page_enhance_scan drives it headlessly. Not yet an option of Recognize text. Refs #4119. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>