What is Visual Zone OCR?

Visual Zone OCR, sometimes just called Visual OCR, is optical character recognition applied specifically to the human-readable portion of an identity document, the printed name, address, date of birth, document number, and other fields laid out in whatever format the issuing country chose to use.

This is a distinctly harder recognition task than MRZ OCR, precisely because there's no single standardized font, layout, or character set governing the visual zone the way ICAO 9303 governs the MRZ. A document reader doing visual zone OCR needs to handle dozens of languages and scripts (Latin, Cyrillic, Arabic, Hebrew, Greek, and others depending on document coverage), varying fonts and type sizes, and field layouts that differ not just between countries but often between different generations of the same country's documents as newer, redesigned versions get issued over time.

To manage this complexity, most mature visual zone OCR systems rely on a large database of document templates, essentially a reference library that tells the software, for a given document type, where the name field typically sits, what font it's likely printed in, and how to distinguish it from surrounding fields like the address or issuing authority. Without this kind of template-based approach, visual zone OCR would need to essentially solve general-purpose text recognition from scratch for every new document type it encounters, both slower and less accurate.

Visual zone OCR also has to handle transliteration in many cases, converting a name printed in a non-Latin script into a Latin-character representation that preserves the sound of the name as closely as possible, since many downstream systems (databases, forms) expect Latin-character input regardless of the document's original script.

Because it's solving a genuinely harder problem than MRZ OCR, visual zone OCR benefits enormously from a large, well-maintained document template library and continuous updates as new document versions are issued globally.

ScanDoc's visual zone OCR draws on a document template database covering a very wide range of countries and document types, and is designed to work in combination with MRZ extraction so the two sources validate each other rather than either one operating in isolation.