Blur, glare, a tilted page, faded ink and heavy compression each cause different text-recognition mistakes. A stronger filter can make the preview look sharper while it deletes punctuation or changes a letter or digit.

Step by step

  1. Work out what’s wrong.Is the page blurry, faint, cut off, grainy, shadowed or warped? The cause tells you whether to retake the photo or clean up the one you have.
  2. Retake it if you still have the paper.Fix the light, distance, steadiness and crop before you spend time correcting a copy.
  3. Clean up gently.Straighten the page and adjust contrast, but keep an unfiltered original. Compare small text and punctuation after each change.
  4. Say when you can’t read it.If key words are still unclear, mark them to be typed by hand or ask for a better copy. Don’t send text you aren’t sure is right.

Check before you send

Search for words you know are on the page and compare the text letter by letter. A page can be searchable enough to find it but not accurate enough to copy names, amounts or legal terms from.

  • You know what’s wrong with the scan.
  • You tried a retake if you could.
  • You still have the unfiltered original.
  • Unclear values are marked for someone to check.
In Paperlane

Paperlane can help you get a clearer scan for OCR, but it can’t guarantee correct text from a damaged page.

How to tell if you can fix a poor scan

Bad OCR results have a cause: blur, glare, a tilted page, low contrast, faded ink, heavy compression, or a crop that cut off letters. Find the cause before you pick a filter or run OCR again.

If you still have the paper, retaking the photo is usually the best fix. If you don’t, keep the current image, make a lightly cleaned copy, and mark any names, numbers or legal phrases you can’t read clearly. Don’t guess them.

How to decide

Retake the page or ask for a better copy when an important field is still unclear after one cleanup attempt. Being able to search the file helps you find it, but it doesn’t prove every character is correct.

Try this first

Check a heading, a paragraph of small text and a hard-to-read number. Stop when more processing changes the page more than it makes it easier to read, and keep the original image.

Example

Say the date on a blurry page is almost readable, but one digit is fuzzy. Sharpening can make the edges look crisper, but it can’t tell you which digit is really there.

If you still have the paper, retake the photo. If you don’t, mark the field for someone to check by hand, and keep the image with any corrected text.

Questions people ask

Should I use aggressive black-and-white?

Only after you’ve kept the original and checked that faint marks and punctuation are still visible. Strong filters can hide details you need.

When should I stop trying OCR?

When an important field is still unclear after you’ve retaken and cleaned up the page. Checking it by hand is safer.

When should I stop trying to improve OCR?

Stop when the scan doesn’t have enough detail to tell the characters apart. Typing the text by hand or getting a better copy is safer than trusting text that may be wrong.

Use recognized text to find documents, not as a replacement for the page. Compare names, numbers and layout with the scan before you reuse or share the text.