censor face and data in an image
Censoring a face or a document in a photo has three common techniques: a solid fill, pixelation, and blur. Only one of them actually destroys the original information; the other two are obfuscation, not erasure, and can be reversed under certain conditions.
There is no "intensity" in blackout, it is the only technique where the result does not depend on a strength parameter.
Intensity 100 would reduce the region to 100 px blocks (just 3 blocks total), while 0 would use 13 px blocks.
The radius never exceeds the mirrored margin the tool uses as a sample, so the blur never pulls colour in from outside the marked region.
Blackout is the only genuinely safe one: it replaces the region with a solid colour, destroying the pixels outright. Pixelation and blur merely scramble the information and, at low strength, can be partially reversed. For faces, documents and license plates, use blackout, which is why it is the tool's default.
It depends on the intensity and the original size of the face: large enough blocks (high intensity) make reconstruction impractical, but pixelation is never a mathematical guarantee the way blackout is. For an identity that absolutely cannot leak, use blackout.
Because the blur radius is a fraction of the shorter side of the marked region itself, not an absolute value. A small region gets a small radius in absolute pixels, but it covers a larger slice of the total area, which is why it looks stronger.
Yes, each marked region has its own mode and its own intensity, so it is possible to apply blackout to a license plate and blur to two faces in the same pass, without downloading and reprocessing the image between markings.
No: the sample used for the blur is mirrored from inside the region itself before the filter runs, and the maximum radius is capped to that mirrored margin, so the result never reads pixels from outside the marked rectangle.
Censoring is applied entirely in your browser. No image is sent to any server, essential for sensitive content.