Both OCR (short for optical character recognition) and ICR (short for intelligent character recognition) are solutions for extracting data from images. Any type of document can be handled with these two technologies, which basically read the document and convert images to processable electronic data units.
What are the pros and cons of OCR?
OCR is a reliable technology, which is usually used for extracting data from printed documents. Since printed documents involve mostly uniform fonts and characters, when you scan the document OCR can easily analyze the dark and light patterns and understand which characters are used. So that it can turn images of each single letter and number into text. Though, OCR is not very handy to process handwriting or noisy images. Because OCR systems need well defined rules to match dark and light patterns with correct characters (letters & numbers), non-uniform contents such as handwriting, shapes, tables, lines, QR codes are out of its scope.
How does ICR work?
ICR is the advanced version of OCR, which means that ICR is capable of recognizing non-uniform characters. It is especially advantageous to use ICR to handle handwriting. Think about tones of handwritten documents that need to be processed manually: forms, invoices, receipts, delivery notes, registration documents etc. It takes huge amounts of time to process them manually and manual data entry often leads to critical errors.
Machine learning is the key feature of ICR, which makes the difference. Machine learning abilities (especially utilizing neural networks) let an ICR system to learn by itself and interpret images without applying to pre-defined rules or templates. This method is called cognitive data capture, meaning extracting data from any type of document by understanding the context of the document and comparing it with many other variations.
Comparison of OCR and ICR
- OCR systems are template or rule based and don’t use AI. That’s why OCR needs human supervision frequently. On the contrary, ICR warns only when an anomaly occurs.
- While OCR is useful for companies who process documents with fixed layouts, ICR is adaptive and trained for frequent layout changes.
- Templates, rules or layouts have to be manually created for OCR. ICR doesn’t require templates.
- Outputs of ICR systems are more easily integrated to ERP systems
- Accuracy rates of OCR are dependent of supporting data base. ICR improves its own accuracy level by time.