JPG or PNG for scanned documents

What the formats actually do

JPG compresses by throwing away detail the eye is unlikely to notice, especially in smooth gradients. On a photograph that trade is invisible. On black text against white paper it produces faint grey fringing around every letter, which gets worse each time the file is saved again.

PNG stores pixels losslessly. A scanned page of text stays crisp, a screenshot of a table keeps its edges, and repeated saves do not degrade it. The cost is size: a PNG of a photographic page can be ten times larger than the same page as a JPG.

Sizing it in practice

A typical A4 page scanned at 300 DPI is roughly 2,500 by 3,500 pixels. As a high-quality JPG that lands around 1–3 MB depending on content; as a PNG it can be 8–15 MB for a photographic page and much less for a page of plain text.

Resolution matters more than format for legibility. A 150 DPI JPG of a typed page is readable; a 600 DPI JPG of a blurry photograph is not. If the page will be printed, aim for 300 DPI; if it will be read on a screen, 150 is enough.

Format and text recognition

OCR works from the pixels, so a clean 300 DPI scan recognises better than a heavily compressed one. JPG artefacts around letters cost accuracy at low quality settings, which is why archival scanning projects use lossless formats and convert afterwards.

If you only need the text, scan once at 300 DPI into PNG, run recognition, and keep the text rather than the image.

Converting between them

Converting PNG to JPG shrinks the file but cannot recover detail that was never lost — it only avoids adding more. Converting JPG to PNG makes the file larger without restoring anything, which is rarely worth doing.

When you build a PDF from images, the images are embedded as they are. Converting to JPG first will make the PDF smaller; keeping PNG will keep text sharper at the cost of size.

• Screenshots and line art: PNG.
• Photographs and camera scans: JPG at high quality.
• Anything destined for OCR: 300 DPI, lossless if possible.