Concept & Knowledge Hub
Document Scanning: Perspective Correction & Binarization
Document Scanning transforms an angled, shadowed photograph of a piece of paper into a flat, perfectly legible, high contrast digital document using advanced image processing algorithms.
Using commercial scanning apps often forces you to upload your sensitive contracts or IDs to third party servers. This tool utilizes OpenCV.js to process the mathematical matrices entirely client side.
Core Architecture & Mathematical Formula
Clean Scan = Edge Detection ➔ Perspective Warp ➔ Grayscale Conversion ➔ Adaptive Thresholding
The scanner first uses the Canny Edge Detection algorithm to find the 4 corners of the paper. It then applies a perspective warp to flatten the image, and finally uses adaptive thresholding to turn gray shadows into pure white.
Best Practices & Essential Guidelines
- Ensure High Contrast Backgrounds: A white piece of paper on a white table is a mathematical nightmare for edge detection. Always place the document on a dark table so the algorithm can instantly find the 4 corners.
- Avoid Harsh Glare: If you scan glossy paper under a bright light, the glare will blow out the pixels to pure white, completely destroying the underlying text. Use diffused, indirect lighting.
- Use B&W Binarization for Text: If the document is purely text (like a contract), always use the 'Black & White' filter. It mathematically removes all colored noise and shadows, drastically reducing the final PDF file size.
Frequently Asked Questions (FAQ)
Why did the scanner warp my document incorrectly?
If the algorithm selects the wrong 4 corners, the resulting warp will severely distort the image. Always use the manual corner adjustment handles to ensure the crop box perfectly aligns with the paper.
Can I scan multiple pages into one file?
Yes. Professional scanning workflows allow you to process multiple images sequentially and batch them into a single, multi page PDF document.
Does this tool read the text?
No. A scanner simply creates a flat image. To extract the actual text strings for editing, you must pass the resulting scan through an Optical Character Recognition (OCR) engine.