OCR (Optical Character Recognition)

What is OCR?

OCR stands for optical character recognition. It is the electronic or mechanical conversion of images of typed, printed or handwritten text into machine-encoded text. A scan, photo or image-only PDF becomes searchable, editable text that a computer can process further. Unlike a plain photo, the resulting text can be copied, searched and processed automatically.

How does character recognition work?

The process typically runs in three stages. First the image is pre-processed — for example by de-skewing, despeckling and converting it to black and white (binarisation). Next the system recognises the characters, classically through matrix matching (pixel-by-pixel comparison) or feature extraction, which analyses the defining parts of a letter. Finally, post-processing with lexicons and contextual analysis improves accuracy. Modern systems also use neural networks. Accuracy depends heavily on image quality, typeface and language; clean, high-contrast originals yield far better results than blurred photos.

Where does the technology come from?

OCR is older than the computer. As early as 1914, the physicist Emanuel Goldberg built a machine that read characters and converted them into telegraph code. In 1931 Goldberg received a US patent that later passed to IBM; in 1959 IBM coined the term optical character recognition with a system of its own. In the 1970s, Ray Kurzweil developed the first omni-font system, able to recognise text in almost any typeface; in 1976 he presented a reading machine for blind people. Since the 2000s, OCR has been widely available as a cloud and mobile service.

Where is OCR used?

OCR digitises archives, invoices, passports and forms and extracts structured data from them. It should be distinguished from intelligent character recognition (ICR), which targets handwriting. Today OCR often works together with layout analysis so that tables and columns are captured correctly. For learning, OCR is the crucial first step in making analogue material usable: only once a photographed script or a scanned book exists as text can AI tools turn it into summaries, flashcards or audio formats.