vimarsh244/scanclean

ScanClean cleans decoded scanned-page pixels conservatively, preserving text while removing document defects. Its primary workflow is either a Python caller supplying an image or the desktop CLI decoding PDF pages; both invoke the core cleaner, which analyzes and removes defects before returning cleaned pixels and optional audit data. The CLI then writes cleaned PDFs. Documented options include deskewing, text-detector guidance, transparency, bilevel output, and audit overlays. The bundled ONNX detector is documented, but its core integration call is not visible in the partial excerpt; that link is therefore shown as unverified.