How to Fix a Blurry Scanned PDF (Before and After OCR)

The portal wants a legible passport page. The OCR engine wants clean character edges. Your scan looks like it was taken from a moving train. Before you download another "AI PDF enhancer," separate what software can fix from what only a camera can.

What blur actually is, in a scan

A scanned page is a photo with pretensions. Blur means the character edges were soft at capture: wrong focus distance, motion, dirty platen glass, too-low DPI, or aggressive JPEG compression on the phone side. Post-processing can improve contrast and remove noise; it cannot re-create edge detail that never hit the sensor. Any tool promising otherwise is upscaling — inventing pixels, not recovering them.

Decision tree

1. Original still in hand? → Rescan. Clean glass, parallel camera, even light, ~300 DPI (capture guide, DPI question). This is always the cheapest complete fix.
2. Original gone, text mostly readable when you squint? → Cleanup pass: deskew, contrast stretch, despeckle in any image-aware editor. Then OCR and proofread — soft glyphs are where rn/m collapses live (accuracy ladder).
3. Text unreadable without enhancement heroics? → You're out of software runway. Request a better copy, rescan from a printout if you have one, or capture again with the document flat and lit evenly.
4. Zoom-and-print critical (IDs, MRZ, fine print)? → Enhancement is triage only; plan a proper capture before submitting (passport/ID scanning).

What "enhance" tools really do

Reputable enhance features (Acrobat's scan cleanup, image editors' auto-levels) do three honest things:

• Contrast normalization — faded gray-on-gray → darker text on whiter ground. Biggest real win for both eyes and OCR.
• Deskew and border crop — straighten the geometry so lines read as lines.
• Noise/despeckle — remove speckles without eating thin strokes (tune carefully; aggressive settings eat periods and accents).

They generally don't: raise true optical resolution, decode a motion-blurred capture, or fix a photo taken at an angle with page curvature. Marketing blur lines on those claims accordingly.

OCR ordering with a soft scan

• Straighten first — wrong line order destroys more recognition than mild blur (straighten FAQ).
• Contrast before OCR; save aesthetic sharpening for optional cleanup.
• Run OCR, then proofread names, numbers, dates — treat output as a first draft (pipeline).
• If the doc still hides secrets, remember: a clearer text layer makes redaction more possible and more necessary.

Compression interaction (don't make it worse)

Aggressive compression after a soft scan compounds the mush — re-encoded images lose what thin edge contrast remained. If size matters, compress once at a moderate setting and verify legibility at 200% zoom before sending (size checks). When the destination has a hard cap and the scan is already soft, you're at the legibility-vs-limit cutoff — change the submission plan, not the quality dial.