How do I make a grey, speckled scan readable?

A scanned page arrives looking like a photocopy of a photocopy: grey where it should be white, soft where it should be sharp, and dotted with specks that were never on the paper. It is legible if you concentrate. It is not something you want to forward to a client.

Why it looks like that

The scanner did not fail. It recorded what it saw, faithfully, including things your eye edits out when you look at the paper itself.

Paper is not white. It is pale grey, or cream, and under a scanner lamp it stays pale grey. Your brain corrects for that when you hold the page; the file has no brain.

Ink is not black. Printer toner is dark grey, and a second-generation photocopy is greyer still. So a scan often has dark-grey text on light-grey paper — a small difference in tone, which reads as washed out.

Specks are real. Dust on the glass, paper fibres, toner scatter. At scanning resolution they become visible dots.

The fix is not to make the image brighter. Brightness moves the text and the paper together and changes nothing about the distance between them. What you want is to push them apart.

What each control does

• Contrast widens the gap between light and dark. This is the one that fixes a washed-out page, and it is the first thing to reach for.
• Black and white takes the decision to its conclusion: every pixel becomes one or the other. Grey paper becomes white, grey text becomes black, and most speckle disappears with the grey it was hiding in. On an ordinary printed document this is usually the single best thing you can do.
• Greyscale keeps the tones but drops the colour. Useful when a page has a coloured background you want gone but contains photographs you do not want reduced to dots.
• Sharpen tightens soft edges. Helpful on a slightly out-of-focus phone photograph; on an already-crisp scan it mostly amplifies the speckle.
• Brightness is for genuinely under- or over-exposed pages, and is the last thing to touch, not the first.

Enhance Scan applies these to every page of a PDF, or to a single photograph.

The thing nobody mentions

Cleaning a PDF turns its pages into pictures.

The adjustments above are image work — contrast, thresholding, de-speckling — and to do image work to a page you have to draw the page as an image. What comes out is a PDF of cleaned pictures.

For a scan this costs you nothing at all. A scan was already a picture of words; a cleaner picture of the same words is pure gain, and recognition works better on it afterwards.

For a document that had real text in it, it is a genuine loss: the words stay legible and stop being selectable, searchable or copyable. That is why you are warned when the file you uploaded still has a text layer — it is a decision worth making deliberately rather than discovering later.

Clean first, then recognise

If your goal is to get the words out, or to edit them, do the cleaning first.

Recognition starts by deciding which pixels are ink and which are paper. On a grey, speckled page that decision is made badly, and every later stage inherits the mistake. A thresholded black-and-white page hands it an answer instead of a guess.

The order that works: enhance, check it looks right, then run OCR PDF or extract the text.

It usually gets smaller too

A black-and-white page compresses far better than a grey one, because there are two values to store instead of two hundred and fifty-six. A cleaned bundle of scans is commonly a fraction of the original size, which is a pleasant side effect rather than the point. If size is the actual goal, Compress PDF targets it directly.