Compressing a scanned PDF

Every other PDF guide tells you to try a lossless pass first. For a scan, that advice is nearly useless, and it is worth understanding why before you spend time on it.

A scanned document contains no text. It is a photograph of a page. There is no text layer to preserve, no fonts to subset, and usually no revision history to discard — so rewriting the file's internal structure saves a few percent at best. The images are the document, and the images are the whole file.

That means compressing a scan is always a resolution and quality decision. The good news is that scans are usually enormously over-resolved, so the first large saving is close to free.

Why your scan is so large

Scanner defaults are set for archival, not for sending:

Setting	Megapixels per A4 page	10 pages, colour
600 dpi colour	35 MP	60–120MB
300 dpi colour	8.7 MP	15–30MB
300 dpi greyscale	8.7 MP	5–10MB
200 dpi greyscale	3.9 MP	2–4MB
150 dpi greyscale	2.2 MP	1–2MB

A page of printed text does not contain 35 megapixels of information. It contains maybe two thousand glyphs. Scanning at 600dpi in colour stores the paper's texture, the scanner's sensor noise, and the faint yellow of the page, in exquisite detail.

The two free wins

Greyscale, if it is black ink on white paper. A colour scan of a monochrome document spends roughly three times the bytes to record that the paper is slightly cream and the ink is slightly blue. Converting to greyscale typically removes 60–70% of the file with no loss of anything you care about. Only keep colour if the document has colour that matters — a stamp, a signature in blue ink that must be provably blue, a highlighted passage.

Lower the DPI to what is actually legible. 300dpi is the standard for scanning because it is the standard for printing. If the document will be read on screen, 150–200dpi is comfortable and is a quarter of the pixels.

Doing both takes a 60MB scan to around 2MB before any lossy compression at all.

Where it gets harder

Text is the worst case for JPEG compression, which is the opposite of the situation for photographs.

Sharp black-on-white edges are the highest-frequency content there is, and JPEG discards high-frequency detail first. The artefacts show up as grey fringing and mosquito noise around every letter, and on small type they can make a document genuinely harder to read rather than just softer-looking.

So the priority order for a scan is the reverse of the one for a photo:

• Keep pixels, drop quality — up to a point. Text needs resolution to stay legible more than it needs precision in each pixel. A 200dpi page at JPEG quality 0.6 reads better than a 100dpi page at quality 0.9.
• But do not drop quality below about 0.45. Past that the fringing merges letters together and a scan of 8pt type becomes guesswork.

That is why a sensible tool searches resolution first, holding quality high, and only steps quality down when it has run out of resolution. Blur from too few pixels and blur from a hard quality setting look different, and running out of pixels is the uglier failure.

The DPI floor, by type size

Roughly where legibility breaks for a clean scan of printed text:

Body text	Comfortable	Readable	Struggling
12pt and larger	150 dpi	100 dpi	72 dpi
10–11pt	200 dpi	150 dpi	100 dpi
8–9pt	300 dpi	200 dpi	150 dpi
Handwriting	300 dpi	200 dpi	150 dpi

Handwriting and poor-quality originals need more than printed text, because the edges are already soft and compression artefacts have less to hold on to.

The thing worth doing instead, if you can

Re-scan at a sensible setting. If the original document is still on hand, scanning once at 200dpi greyscale is better in every way than compressing a 600dpi colour scan down to the same size — you avoid a lossy pass entirely, and the scanner's own downsampling is better than any after-the-fact resize.

Or run OCR. If your scanner or PDF software can add a text layer, do it before compressing. The document stays the same size but becomes searchable, and the text layer survives even if the images are later compressed hard. It does not make the file smaller; it makes a smaller file much more useful.

Hitting a form's number

Scans are mostly compressed because a portal demanded a size. The common targets are 200KB, 300KB and 500KB, and they are tight for a multi-page scan.

The honest arithmetic: a 200KB budget across 10 pages is 20KB per page. At that budget you are looking at roughly 100dpi greyscale at moderate quality — readable for 12pt print, marginal for anything smaller. If the form wants 200KB and your scan is 20 pages of small type, the right answer may be to split the submission or re-scan rather than to compress harder.

A tool that measures rather than estimates will at least tell you honestly where it landed:

Lossless rewrite only            14.2MB   over 500KB
scale 2.0 (144dpi) quality 0.78   6.8MB   over
scale 1.2 (86dpi)  quality 0.78   1.1MB   over
scale 0.8 (58dpi)  quality 0.78   520KB   over, barely
scale 0.75 (54dpi) quality 0.78   480KB   under, best found

Compress a PDF to 200KB, to 500KB or to 1MB tries the lossless pass first regardless — it costs one measurement and occasionally a scan does have save bloat to shed — then searches resolution before quality. It runs in your browser, which for scanned passports, bank statements, medical records and signed contracts is the part that matters most.

FAQ

Why does compressing a scanned PDF lose quality when compressing a normal PDF does not?

Because a scan has no text layer to protect. A text-based PDF can be made smaller by rewriting its internal structure — compressing object streams, dropping revision history — with every glyph still a glyph afterwards. A scan is images of pages, so the only way to make it smaller is to store those images with fewer pixels or less precision, and both are visible.

What DPI should I scan at to keep the file small?

200dpi greyscale for ordinary printed documents you will read on screen, which is comfortable for 10pt type and a quarter of the data of 300dpi colour. Use 300dpi for small type, handwriting, or anything that will be printed. Scanning in colour when the document is black ink on white paper roughly triples the file for nothing.

Can I make a scanned PDF searchable while compressing it?

Not in the same step, and the order matters: run OCR first, then compress. OCR adds an invisible text layer that survives image compression, so the document stays searchable even after the pictures are compressed hard. Doing it the other way round means OCR runs on degraded images and reads worse.

How small can a 10-page scan get and stay readable?

Around 300–500KB for 12pt printed text in greyscale, which works out to 30–50KB per page at roughly 120–150dpi. Below about 20KB per page you are under 100dpi and body text starts to break down. If a form demands 200KB for a 20-page scan of small type, compressing to fit will produce something that is technically accepted and practically illegible — splitting the submission or re-scanning is the better answer.