How to redact a scanned document

A scanned contract, a photographed form, a faxed statement: to a computer, each page is a picture. You can't search it, and a tool that only looks for text will find nothing to redact.

A scan is a picture of text

When you scan a page, the result is an image, even when it is saved as a PDF. The names and numbers are dots of colour, not letters. To find them automatically, a program first has to read the image with text recognition (often called OCR, for optical character recognition), which turns the shapes back into words and notes where each word sits on the page.

Where text recognition struggles

Text recognition works well on clean, straight, printed pages. It has more trouble with:

• low-resolution or blurry scans, and photos taken at an angle;
• crooked or cropped pages;
• faint photocopies and coloured or patterned backgrounds;
• handwriting, stamps and signatures;
• very small print, and forms with many boxes and lines.

When a word is misread, a program looking for a name or a number may not recognise it. On scans, the page-by-page review matters even more.

Two traps specific to scans

• The invisible text layer. Many "searchable" scans carry hidden text on top of the image. If you black out the image but leave that text, the name can still be copied or found by search. If you remove the text but not the image, the name is still visible.
• Covering instead of changing the image. A box drawn over a scan is a separate object on top of the picture. Unless the image itself is changed, the original can remain in the file underneath.

A safe way to redact a scan

1. Scan cleanly. A resolution of 300 dpi is a common setting for text; keep pages straight and complete.
2. Read the scan on your computer rather than uploading it to an online service.
3. Review each page at a zoom level where you can read the small print. Check headers, footers, stamps and handwritten notes.
4. Search the recognised text for names and numbers you know are in the document.
5. Export a new copy in which the hidden areas are part of the image itself, with no leftover text layer underneath.
6. Open the exported copy and check it again.

How Anonyma handles scans

Anonyma reads scanned pages on your computer with text recognition; nothing is uploaded. In the exported copy, scanned pages are rebuilt as images, so the hidden areas are part of the picture, and after export Anonyma reads the copy again to confirm that the hidden details are gone. The copy is a new file, and your original is not changed.

Text recognition can still misread a poor scan, and automatic detection can miss things. You review every page before exporting, and you can search for any word or add your own boxes.

Related: why a black box isn't enough and how to redact PDFs without uploading them.