Arabic OCR

Optical Character Recognition

OCR is the ability to scan a document (or grab a PDF file) and run an OCR program on it and it will generate, based on optical recognition and approximation, an editable text file. For an idea about OCR see http://www.students.cs.uu.nl/people/mjkammer/Work/intro_2_OCR.html

Current Status of Open Source Arabic OCR software

The only FOSS OCR system with Arabic support is Tesseract, help is needed in testing and training it.

Resources

• OCR from Wikipedia

Arabic OCR Links

Papers

• Automatic Recognition Using Zernike Moments As A Feature Extractor (Paper)
• Graph Based Segmentation .. (Paper)
• Structural Features Of Cursive Arabic Scripts (Paper)
• Multilingual Machine Printed OCR (Paper)
• Test of two Arabic OCR programs
• Performance Evaluation of two Arabic OCR products

Software (FOSS)

• Tesseract is an open source OCR, initially developed by HP, and released under the Apache License. 3.x versions has Arabic support.
• GOCR - included in Debian and other distributions. No Arabic support.
• GNU Ocrad "is an OCR [...] program based on a feature extraction method". No Arabic support.

Other Links

• How to encode image produced by a recognition system (mailing thread) http://lists.arabeyes.org/archives/general/2002/March/msg00001.html
• Rapidly Retargetable Translingual Detection http://tides.umiacs.umd.edu/description.html