Blur or Pixelate Image

Why the method you pick matters

The three options here are not interchangeable, and the difference is a real security property rather than an aesthetic preference.

Solid black is absolute. The pixels are replaced with black, and there is no information left in the region to recover by any means. It looks blunt, and that bluntness is exactly the guarantee.

Pixelation at a reasonable block size is effectively irreversible too. Each block is replaced by the average colour of the pixels it covered, and averaging destroys information permanently. The caveat is block size: pixelating text into very large blocks is safe, while a light pixelation with many small blocks retains enough structure that the original can sometimes be reconstructed, particularly for text drawn from a known alphabet.

Blur is the weakest of the three and the most commonly misused. A Gaussian blur is a reversible mathematical operation in principle, and mild blurs have been successfully undone by sharpening. Blurred text and blurred faces have both been recovered in published research. Blur looks the most professional and protects the least.

The failure mode worth knowing about

In 2016 researchers demonstrated that neural networks could identify faces and read text through both blurring and mosaicing at the settings commonly used by broadcasters and publishers. The technique works because obscuring a region does not remove all correlation with the original; it just makes the correlation harder for a human to see.

The practical lesson is not that pixelation never works — a heavy mosaic destroys genuinely enough information — but that light obscuring is theatre. If a region is being hidden because disclosure would cause harm, use solid black or pixelation at a large block size, and judge it by whether you can still make out shapes rather than by whether it looks tidy.

A second failure mode is more mundane and more common: obscuring the wrong thing. A screenshot with the account number blacked out but the sort code visible, or a photo with the face blurred but a name badge readable, protects nothing. Look at the whole frame, not just the obvious element.

Common jobs this handles

Screenshots for documentation and bug reports are the most frequent use: hiding email addresses, API keys, customer names and internal URLs before posting to a tracker or a forum. This is worth doing habitually, because screenshots posted publicly are indexed and archived.

Photographs shared online where other people appear incidentally — children in the background of a holiday photo, faces in a crowd, a colleague at a neighbouring desk. Those people did not consent to being posted.

Documents photographed for a claim or an application, where the recipient needs to see one part and has no business seeing the rest. And vehicles: a number plate identifies a registered keeper in most countries.

For property photos, remember that house numbers, distinctive doors and street signs are all identifying — and so is the metadata, which is a separate tool on this site.

Working with the tool

Boxes can be layered and overlapped freely, and the preview updates as you draw so you can judge the result at actual size rather than guessing. Zoom your browser in if you need finer placement on a large image.

The output is always PNG, which is lossless. That matters here: saving a censored image as a heavily compressed JPEG can produce artefacts at the boundary of a redacted region, and in extreme cases those artefacts carry a faint trace of what was underneath.

One thing to be careful about is the source. Censoring a copy and keeping the original is the right workflow, but make sure it is the censored copy you actually send. The most common failure in practice is not a weak blur — it is attaching the wrong file.