Prepare for OCR

POST /api/document-prepare-for-ocr

Prepare a document scan for OCR: deskew, binarize, denoise, contrast-stretch. Improves recognition rates on noisy phone-camera captures and skewed scans.

Cleans up a document image so the OCR engine has the best possible chance. Apply this before `document-ocr` for noisy inputs — phone-camera captures, photographed receipts, faxed pages. The pipeline is: - Deskew (auto-detect rotation up to ±15° and straighten). - Convert to greyscale. - Adaptive threshold to clean black-on-white binary. - Despeckle (remove isolated stray pixels). Output is a PNG ready to feed straight into `document-ocr`.

Inputs

Name	Type	Default	Description
image*	file	—	Input document image.
deskew	boolean	true	Rotate to correct skew.
maxAngle	number (0…90)	15	Cap on detected skew magnitude (degrees).
thresholdMode	enum (otsu \| adaptive)	"otsu"	Binarisation strategy.
denoise	boolean	true	Light Gaussian blur before threshold.

Response

Modes: binary, json. Cache: yes (24h TTL).