Quick Concepts: Optical Character Recognition
What is Optical Character Recognition?
Optical character recognition (OCR) is the process of converting images of text into machine-readable text. OCR can extract characters and words from static PDFs, photographs, and other images and convert them into meaningful text that can be searched, edited, organized, and analyzed. This is done via a combination of hardware (such as optical scanners) and software. With the addition of artificial intelligence, OCR can recognize multiple languages and read handwritten words in different writing styles.
How does OCR work?
Starting from a photo or scanned physical document, the OCR software first converts it into a black-and-white or two-color image, using the darker color for text characters and the lighter color for the background. Next, one of the two following algorithms is used to translate the text images into characters:
- Pattern recognition – Whole characters are extracted and compared to examples of text that have been fed into the program. This is effective for typewritten text in commonly used fonts.
- Feature detection – Characters are broken down into features such as angled lines, curved lines, loops, intersections, etc., and compared to an abstract character representation in the program. This method is better at recognizing unfamiliar fonts or handwritten text.
How is OCR used?
OCR has a variety of applications, often operating behind the scenes or embedded in other functions. Use cases for OCR include extracting information from static documents and unstructured data, automating data entry, digitizing physical documents into searchable formats, reading passports and IDs at airports, assisting the visually impaired, interpreting traffic signs, scanning vehicle license plates, depositing checks electronically, and performing language translation on text within images. Many of the automated products and services that we use today are powered or supported by OCR.
Accelerate AI with Annotated Data
Check Out this Article on Why Your Model Performance Problems Are Likely in the Data