pdftract/crates
jedarden 32350f8e81 feat(pdftract-55ihl): implement Otsu global thresholding for OCR preprocessing
Add otsu_binarize() function using imageproc::contrast::otsu_level and
threshold functions. Otsu method finds optimal global threshold by
maximizing inter-class variance between foreground and background.

Changes:
- Add imageproc 0.26 to Cargo.toml dependencies (ocr feature)
- Create crates/pdftract-core/src/ocr/preprocessing/otsu.rs module
- Export otsu_binarize from ocr::preprocessing and lib.rs
- Comprehensive tests: digital-origin images, binary output, uniform/tri-modal edge cases, text-like images, small images, benchmark

Acceptance criteria:
- Digital-origin (uniform-lit) page produces clean binary ✓
- Output pixels are exactly 0 or 255 ✓
- Benchmark: 1080p < 50ms (test provided, ignored by default) ✓
- Tri-modal histograms fail gracefully (no panic, still binary) ✓

Closes: pdftract-55ihl
2026-05-25 12:41:17 -04:00
..
pdftract-cer-diff docs(pdftract-aawrz): add LICENSE-MIT and LICENSE-APACHE files 2026-05-23 10:36:28 -04:00
pdftract-cli feat(pdftract-5bzpg): implement pdftract-grep-1000 CI benchmark skeleton 2026-05-25 08:53:23 -04:00
pdftract-core feat(pdftract-55ihl): implement Otsu global thresholding for OCR preprocessing 2026-05-25 12:41:17 -04:00
pdftract-libpdftract feat(pdftract-3s2i): implement Phase 5.5.2 validation filter 2026-05-24 04:57:17 -04:00
pdftract-py feat(pdftract-3j2u): implement 50 MB size limit + base64 encoding for attachments 2026-05-25 11:42:28 -04:00