Guide

THE TOOLBOX / DOCUMENT VISION & OCR PREP

Clean scans for OCR,
Otsu binarized locally.

Remove paper shadows, boost faded ink, and invert dark-mode screenshots into crisp monochrome OCR masks right in your browser—with zero server uploads.

1. Document Image & Otsu Binarizer

Drop a document photo or screenshot (or paste Ctrl+V)

Applies local Rec. 601 luminance & Otsu thresholding for crisp monochrome OCR masks—zero server uploads.

1.30×

2. Transcription & Text Pad

0 words0 chars0 lines

PREPROCESSING PRESETS

Recommended OCR cleanup settings by document type.

PresetContrastInvertBest Used For
Standard Document1.25×OffPrinted pages, PDFs, and clear book scans
Faded Receipt1.85×OffThermal store receipts and low-ink invoices
Dark Mode Code1.40×OnIDE screenshots, terminals, and dark UI captures
Pencil / Handwriting2.20×OffLight graphite notes, whiteboards, and sketches

1. Rec. 601 Luminance + Otsu Math

Converts color pixels using perceptual luminance weights before finding the exact histogram threshold (0–255) that separates ink from background shadows.

2. Live Ink Coverage Telemetry

Check the dark-pixel ink percentage immediately—typical clean text scans land between 5% and 25% ink coverage.

3. 100% Private Document Handling

Tax forms, receipts, IDs, and internal screenshots stay completely in local browser memory.

Frequently Asked Questions

What is Otsu thresholding and why does it improve OCR accuracy?

Otsu thresholding analyzes an image’s 256-level brightness histogram and finds the optimal cutoff between dark ink and light paper. Converting shadowed photos into crisp black-and-white masks dramatically reduces OCR character errors.

Are my receipts, invoices, or screenshots uploaded to a server?

No. All grayscale luminance conversion, contrast amplification, and Otsu binarization run 100% locally in your browser using HTML5 Canvas. Zero images leave your device.

Which setting works best for faded thermal receipts or pencil notes?

Use the "Faded Receipt (1.85×)" or "Pencil / Low Contrast (2.2×)" preset. Boosting contrast widens the gap between light gray strokes and paper background before binarizing.

How should I prepare dark-mode code or terminal screenshots for OCR?

Click "Dark Mode Code (Invert)" or check Invert Luminance. Most OCR engines expect black text on a white background, so inverting light-on-dark screenshots improves accuracy immediately.

Can I download the cleaned black-and-white mask for Tesseract or Python OCR?

Yes. Click Download Mask PNG to save the full-resolution Otsu-binarized image for Tesseract CLI, Apple Vision, or OpenCV workflows.

Can I paste a screenshot directly from my clipboard?

Yes. Copy any screenshot and press Ctrl+V (Windows/Linux) or Cmd+V (Mac) anywhere in the studio to binarize it instantly.