OCR: how to recognise the text of a scanned document

▶ 1. What OCR is

OCR stands for Optical Character Recognition: it is the technology that looks at the image of a page and extracts the text from it, letter by letter.

It is needed because, to a computer, a scan is not a document but a photograph. The words you read perfectly well are, for software, merely light and shade: they cannot be selected, searched, copied or corrected.

OCR bridges exactly that gap. After recognition the page contains real text and the document behaves like a normal PDF again.

🔍 2. When you actually need it

Not every PDF needs OCR. There is a simple way to tell in two seconds.

Try selecting a word with the mouse. If the text highlights, the document already contains real text and OCR is pointless. If instead the cursor makes you draw a rectangle as it would on a photo, you are looking at a scan.

• ✓ Paper documents run through a scanner or photographed with a phone.
• ✓ Faxes, receipts and forms archived years ago in image format.
• ✓ PDFs produced by older business systems that print to an image.
• ✓ Screenshots and photos of pages inserted into a PDF.

◎ 3. Running OCR

The editor spots pages without text by itself and offers recognition, but you can also start it whenever you like.

1. Install the extension from the download page; it is a Chrome extension and works the same on Windows, Mac and Linux.
2. Open the document with Choose a PDF (or Open PDF).
3. If the Scanned pages detected window appears, carry on from there; otherwise open Local OCR from the toolbar.
4. Choose the language of the document, or leave Automatic.
5. Click Run local OCR and wait for the progress bar to complete.
6. When it finishes the text is selectable: you can correct it, search it and copy it.

🌐 4. The languages recognised

10 languages are included, all available offline: nothing has to be downloaded at the moment of use.

1. If you know the language of the document, select it: recognition is faster and more accurate.
2. If the document is mixed or you are unsure, leave Automatic: all the included languages are loaded.
3. The automatic option takes more time and more memory, so it is worth using only when you really need it.

⚖ 5. Getting a good result

The quality of OCR depends almost entirely on the quality of the original scan: no engine recovers what is not in the image.

• ✓ Adequate resolution — a scan at 300 DPI is the usual reference for text; below 200 DPI errors increase sharply.
• ✓ A straight page — if the scan is crooked, straighten it first with page rotation.
• ✓ Good contrast — black text on white works better than a faded photocopy or a photo taken in poor light.
• ✓ Printed text — OCR reads typeset characters; handwriting is out of reach.
• ✓ Always proofread — on figures, reference codes and amounts a manual check pays off: a 0 read as O slips by easily.

➔ 6. What you can do after recognition

Once the text exists, the document behaves like any other PDF.

1. Correct and update the text: see the guide on how to edit the text of a PDF.
2. Search for a word across the whole document, which is impossible on a scan.
3. Highlight the important passages, as explained in the guide on how to highlight a PDF.
4. Remove confidential data properly, following the guide on how to redact text.
5. Archive the document as a searchable file, so you can find it later by searching for a word.

? 7. Frequently asked questions

The questions that come up most often on this topic.