unblur guides

Whether a blurry image can be recovered depends on why it is blurry. Slight misfocus can often be improved with an unsharp mask at a 0.8–1.5 px radius, because the edge information is still in the file. Motion blur is harder: the direction and length of the movement must be estimated, and recovery is usually partial. Heavy pixelation or a genuinely low-resolution source cannot be recovered at all — the detail was never recorded. AI enhancers can help with the first category by inferring structure such as faces, foliage and fabric, but they invent plausible texture in the second. That distinction matters most when the unreadable part is a value someone will act on: an account number, a dosage, a plate, a serial. An AI model will happily produce a confident-looking digit that is simply wrong. The guides here set out working parameters, a sensible 2× ceiling on upscaling, the correct order of operations (crop, contrast, then sharpen last), and the cases where the honest answer is to reshoot or request the original file rather than enhance a compressed screenshot.

4 articles in this collection, newest first.

← All articles