Clean up a scanned PDF
Upload the book once and every page comes back with white paper, sharp black print, no specks and straight lines, while pages with pictures stay in greyscale.
Key takeaways
- Cleaning a scan is five jobs: whiten the paper, darken the print, remove specks, straighten tilted pages and trim dark edges.
- Doing it page by page in a PDF editor takes an evening for one chapter. FixMyScan does every page of the book in one pass.
- A contrast slider turns photos and engravings into black blobs. Picture pages are spotted and kept in greyscale instead.
- If you'd rather keep every page in grey, pick Clean greyscale: the paper is still levelled and the page straightened.
- Clean black-and-white pages are also smaller, read better on e-ink and give OCR a fair chance.
The five things a messy scan needs
Most scanned books have more than one problem at once. A page from an 1890s county history might be yellowed, slightly tilted, faint where the type wore thin, and dotted with foxing. Fixing one of those and leaving the rest doesn’t make it readable. Here is what each fix does:
| Problem | What you see | What the clean-up does |
|---|---|---|
| Yellowed or grey paper | Text sits on a tinted, uneven field | Levels the paper to white |
| Faint or thin print | Letters break up at normal size | Turns the print solid black |
| Specks and foxing | Dots and stains across the text | Removes small marks away from letters |
| Crooked page | Lines slope across the screen | Straightens tilts of up to about 5 degrees |
| Dark scanner edges | Black bars round the page | Removed with the margin crop |
Do it for the whole book, not page by page
Desktop PDF editors can do some of this, one page or one setting at a time. That is fine for a two-page letter. For a 350-page book it means hundreds of decisions, and if one setting suits chapter one it often wrecks the plates in chapter six.
FixMyScan works out the clean-up for each page by itself. You make one choice, the crop, and the rest runs across every page in one go.
How to clean a scanned book
- Upload the PDF or DjVu file. Anything up to 1,000 pages.
- Draw the crop. The pages are layered so you see where text sits across the book. One box trims the margins and the dark edges everywhere.
- Leave the rest to the tool. Each page is straightened, the paper turned white, the print turned black and the specks removed. Text is read and hidden under every page as it goes.
- Review. Pages that look blank, crop unusually, read poorly or carry pictures are flagged. Swap any page back to the original.
- Download. You get one PDF, with every page still in its place.
What happens to pictures
This is where most clean-up tools fail. Push the contrast high enough to make faint text black, and a photograph, a map or a steel engraving goes with it into a solid dark mass.
FixMyScan checks every page for areas of smooth mid-tones and colour. A page that has them stays in greyscale: the paper is still levelled to white, so it matches the pages round it, but the picture keeps its shading. Only plain text pages become pure black and white.
If your book is mostly pictures, such as an atlas, a photo history or a field guide, choose Clean greyscale in your settings and every page is treated that way.
Best suited for: old printed books, out-of-copyright downloads, photocopied course packs and books you scanned yourself with a phone or flatbed.
Why black and white is the default
There are three reasons, all practical:
- E-ink shows a limited range of greys. A grey page with dark grey print has little contrast on the screen. Black on white has the most.
- Search gets better. OCR reads a clean black-on-white page with fewer mistakes than a speckled grey one.
- The file gets smaller. Black-and-white pages are stored losslessly at one bit per pixel, far less data than a grey photo of the same page. See PDF too big for Kindle if size is your main problem.
How other tools compare
| Tool | Whitens paper | Straightens | Runs |
|---|---|---|---|
| FixMyScan | Yes, pictures kept grey | Yes | Browser |
| ScanTailor Advanced | Not checked | Yes | Windows, Linux |
| Sejda | Not checked | Yes | Browser, desktop |
| PDF24 | Not checked | During OCR | Browser, Windows |
| iLovePDF | Not checked | Not mentioned | Browser, apps |
“Not checked” means the tool’s own page didn’t say and we haven’t tested it. Sejda works within its free and paid page limits; PDF24 advertises no limits, which we haven’t tested.
ScanTailor Advanced is free and handles whole books if you are on Windows or Linux and happy to work through its stages. It leaves out OCR, so you’d run a second tool for search.
What clean-up can’t fix
A page photographed out of focus stays soft. Text cut off by the edge of the scan isn’t there to recover. Water damage that has wiped out letters leaves gaps. Cleaning makes what was captured easier to read; it doesn’t invent what wasn’t.
Questions
Is there any way to fix a badly scanned PDF book?
Usually, yes, if the words are legible to you. Cleaning can whiten the paper, strengthen faint strokes and straighten pages. It can't bring back text that was never captured, such as a line lost in a blurred photo or cut off at the edge.
Will my illustrations survive?
Yes. Pages with a meaningful area of mid-tones or colour, such as photos, maps or engravings, are kept greyscale with the paper levelled to white, rather than forced to black and white.
Can I keep the whole book in greyscale?
Yes. Choose Clean greyscale in your settings. It is meant for books full of photos or maps.
Can I clean a scan from Internet Archive or Google Books?
Yes. Public-domain downloads from Internet Archive, Google Books and HathiTrust are a common case: grey, heavy files with dark edges.
Do I have to adjust settings for each page?
No. You set one crop for the book. Whitening, darkening, despeckling and straightening run on every page without settings.
What if one page comes out worse?
Unusual pages are flagged after processing. Use the original for any page you don't like and it is swapped back, keeping its searchable text.
What does it cost?
$9.99 a week or $79.99 a year, with up to 1,000 pages a book.