This comparison of optical character recognition software includes: OCR engines, that do the actual character identification Layout analysis software, that divide scanned documents into zones suitable for OCR Graphical interfaces to one or more OCR engines Software development kits that are used to add OCR capabilities to other software (e.g. forms processing applications, document imaging management systems, e-discovery systems, records management solutions)
Name scope="col" scope="col" class="unsortable" scope="col" License scope="col" Windows Mac OS X Linux BSD scope="col" scope="col" Programming language scope="col" Languages Fonts Output formats Notes ABBYY FineReader 1989 16 2023 C/C++ 198 All fonts DOC, DOCX, XLS, XLSX, PPTX, RTF, PDF, HTML, CSV, TXT, ODT, DjVu, EPUB, FB2 ABBYY also supplies SDKs for embedded and mobile devices. Professional, Corporate and Site License Editions for Windows, Express Edition for Mac.AIDA201613.02024 All languages using Latin alphabetMachine and handprinted text, Latin alphabetDOCX, XLSX, PPTX, TXT, CSV, PDF, JSON, XMLAIDA is able to learn how to extract any value from any document, with a single click on a single document. AnyDoc Software 1989 VBScript Works with structured, semi-structured, and unstructured documents. Asprise OCR SDK 1998 15 2015 Java, C#, VB.NET, C/C++/Delphi 20+ Plain text, searchable PDF, XML Java, C#, VB.NET, C/C++/Delphi SDKs for OCR and Barcode recognition on Windows, Linux, Mac OS X and Unix. CuneiForm 1996 1.1 2011 C/C++ 28 Any printed font HTML, hOCR, native, RTF, TeX, TXTDebian manual page for Cuneiform for Linux version 1.1.0 Enterprise-class system, can save text formatting and recognizes complicated tables of any structureE-aksharayan201014RTF, TXT, BRL GOCR 2000 0.52 2018 C 20+ Google Drive OCR or Google Cloud Vision 2015 Browser Browser Browser Unknown 200+ All fonts text Google blog post Microsoft Office Document Imaging Office 2007 2007 Uses OmniPage Microsoft Office OneNote 2007 2011 2007 OCRFeeder 2009-03 0.8.5 2022 Python Features a full user interface and has a command-line tool for automatic operations. Has its own segmentation algorithm but uses system-wide OCR engines like Tesseract or Ocrad Ocrad 0.29 2024 C++ Latin alphabet Command line OCRopus 2007 1.3.3 2017 Python All languages using Latin script (other languages can be trained) Normal Latin script and Fraktur (other scripts can be trained) TXT, hOCR,OCRopus includes the ocropus-hocr tool which produces hOCR from the recognition results. PDFIn combination with the hocr-tools Pluggable framework under active development, used for Google Books OmniPage 1970s 19.2 2015 C/C++, C# 125 Machine and handprinted fonts DOC/DOCX XLS/XLSX PPTX RTF PDF PDF/A Searchable PDF HTML Text XML ePUB MP3 Product of Nuance Communications Puma.NET 2009 C# 28 Any printed font .NET OCR SDK based on Cognitive Technologies' CuneiForm recognition engine. Wraps Puma COM server and provides simplified API for .NET applications ReadSoft 14? Scan, capture and classify business documents such as invoices, forms and purchase orders integrated with business processes. Scantron For working with localized interfaces, corresponding language support is required. SmartScore 1991 10.5.8 2015 For musical scores Tesseract 1985 5.5.0 2024 C++, C 100+Based on count of language training files for version 3.04. Available at the download page. Any printed font Text, ALTO, hOCR, PAGE,Usage explained in the Tesseract Readme and FAQ PDF, others with different user interfacesSuch as ODF with OCRFeeder or the API Developed at HP Labs (1985–1995) and Google (2006–2018 Name License Windows Mac OS X Linux BSD Programming language Languages Fonts Output formats Notes
Evaluation A 2016 analysis of the accuracy and reliability of the OCR packages Google Docs OCR, Tesseract, ABBYY FineReader, and Transym, employing a dataset including 1227 images from 15 different categories concluded Google Docs OCR and ABBYY to be performing better than others.
References