Blur the text in an image

Why a blurred word is weaker than a blurred face

A blur is a known mathematical operation applied to a known alphabet. Faces vary infinitely; the characters in a card number are one of ten shapes in one of a few fonts, at a size the rest of the image tells you. That is a much smaller problem to solve backwards, and tools to solve it exist.

So words default to a solid fill, which has genuinely replaced the pixels rather than rearranged them. You can switch to a blur, and the editor will say what it costs.

Reading the text out

Recognition runs in your browser, on a copy of the picture scaled down for speed. The boxes it reports are scaled back to full size, so what you cover and what you save line up exactly.

It is good at printed text, fair at screenshots, and poor at handwriting or text at a steep angle. Where it misses something, drag a box over it.

What gets recognized as what

Anything that passes a real test is put in its own group so you can decide about it separately: card numbers that satisfy the Luhn checksum, bank accounts that satisfy the IBAN check, passport lines with valid check digits. The rest is simply text.

• Card numbers, checksum-validated rather than pattern-matched
• Bank accounts, national ID numbers and dates of birth
• Street addresses, emails and phone numbers
• API keys and tokens, the sort that leak in a screenshot of a terminal
• Anything else, by dragging a box over it yourself

How to do it

1. 1Open the imageThe words in it are read and grouped.
2. 2Switch categories on or offOr click a single area to change just that.
3. 3Save the copyCovered areas are solid by default, for a reason.

Questions