Pixelate or black out? Hiding personal details safely

How much information pixelation, blurring and blacking out leave in a photo, research showing pixelation can be defeated, and a safe order for hiding licence plates, ID numbers and faces.

Reviewed: 2026-10-07 · Sources checked: 2026-10-07

Is pixelating enough to hide personal information?

For anything that must not be revealed, blacking out is safer than pixelating. Pixelation and blur only mix the original pixels, so rough shapes remain, and research has shown that pixelated and blurred faces and digits can be recognised with deep learning. Blacking out replaces the pixels with a solid colour, so none of the original information is left there.

1. How the three methods differ

Ways of hiding compared
Method	What it does	What is left	Suitable for
Pixelate	Splits the area into blocks and paints each block its average colour	Colour and brightness per block	Softening background people or signs
Blur	Smears pixels into their neighbours	Blurred outlines	Similar to pixelation
Black out	Replaces the pixels with an opaque solid colour	Nothing (solid colour)	Numbers, addresses, accounts, faces — anything that must not be revealed

2. When pixelation is especially risky

Text and numbers: licence plates, ID, account and phone numbers have a known shape and length, so block colours alone can give them away.
Small blocks: if the blocks are smaller than the letters, the outlines stay almost intact.
Several images of the same thing: slightly different pixelated versions together make recovery easier.
A 2016 study (McPherson, Shokri, Shmatikov) showed that deep learning models can recognise many pixelated and blurred faces and digits.

3. A safe order for hiding details

Use “Black out” for anything that must not be revealed, and make the area generously larger than the text.
If the part is not needed at all, cropping it out is even more certain than covering it.
Open the result file and check that everything is hidden.
The original photo is still on your phone or cloud storage — keep it out of what you share.
In a PDF editor, placing a black box on top may leave the text underneath in the file. Hide details in the image first and make the PDF from that image, or use a real redaction feature.

Sajinbaro's mosaic tool saves one flat image, so the original pixels and the edit history are not stored in the file. Mosaic strength is the block size in pixels of the original photo (8, 16, 32 or 64 px).

Sources and rules

Calculation: The size examples in this guide are generated by the same code as the tools on this site.
Defeating Image Obfuscation with Deep Learning (2016): R. McPherson, R. Shokri, V. Shmatikov. Neural networks identified pixelated and blurred faces, objects and handwritten digits (checked 2026-10-07).
How the tool works: The pixelate and black-out descriptions match what Sajinbaro's mosaic tool actually does (block average colour, opaque black).