How to improve scanned PDF quality

Resolution: 300 DPI, and when to go higher

300 DPI is the standard for text and it is the right default. It resolves 8pt type comfortably, gives OCR enough to work with, and keeps files manageable.

Go to 600 DPI for very small print, footnotes, engineering drawings, or anything you expect to enlarge. Going higher than that on ordinary documents mostly buys you file size.

Do not scan below 200 DPI for anything you intend to print or run OCR on. That detail cannot be recovered later.

Colour mode: greyscale is usually right

• Greyscale for plain text and most documents. Smaller than colour, and it keeps the mid-tones that post-processing needs.
• Colour when stamps, signatures or highlighting carry meaning.
• Black & White at capture time: avoid. It applies a threshold with no chance of review, and once a faint stroke has been thresholded away it is gone for good. If you want a two-tone result, capture greyscale and threshold afterwards, where you can see what you are losing.

Compression: the invisible quality setting

Most scanners default to fairly aggressive JPEG compression. It is invisible on a photograph and quite visible on text, where it produces soft edges and blocky artefacts around every letter.

If your scanner offers a quality slider, move it towards the high end for text documents. If it offers JPEG 2000 or lossless options for greyscale, they are worth the extra megabytes for anything you will keep.

Lighting, for phone scans

• Use diffuse light from the side. A single overhead lamp produces the hard shadow gradient that is hardest to correct.
• Do not let your own shadow fall across the page — the most common phone-scan mistake.
• Hold the phone parallel to the page. Angling it introduces perspective distortion on top of the lighting problem.
• Put the page on a dark, matte surface. It helps the camera meter the page instead of the background, and it stops light bouncing back through thin paper.

Scanner hygiene

Clean the glass. Every speck on a flatbed platen appears on every page you scan, and every mark on a sheet-fed scanner's sensor strip appears as a line down the whole document.

On document feeders, watch for skew: worn rollers pull pages crooked, and a crooked page is harder to read and harder to OCR. If pages keep coming out at an angle, the rollers need cleaning or replacing.

Then enhance, then OCR

Once the scan is as good as capture can make it, run the tone correction — that is where enhancing a scanned PDF fits — and run OCR last, on the cleanest version of the page you have. Recognition engines are trained on clean, high-contrast text, and giving them exactly that measurably improves accuracy.