A New optical character recognition algorithm based on string expressions

A NEW OPTICAL CHARACTER RECOGNITION ALGORITHM BASED ON STRING EXPRESSIONS (M.Sc. Thesis) Derya YILMAZ GAZİ UNIVERSITY INSTITUTE OF SCIENCE AND TECHNOLOGY September 1999 ABSTRACT In this thesis, a new character recognition algorithm and database structure was developed for recognition of character's applications (OCR-Optical Character Recognition). The new algorithm; called DECRA, expresses character in skeletion form, pixel groups called "string" and relationships among the strings. DECRA algorithm produces the requested character information in a simple way that doesn't require complicated mathematical expressions. Structure of the database was built according to the character informations that arised from DECRA. The characters in the database were defined by segments and their properties. The character information from the output of DECRA were segmented. This information is compared with the characters defined in the database and recognition is realized. Performance of the algorithm was tested theorietically and practically. The algorithm has worked correctly. Various characters with theorietically different noise level were studied and the algorithm was developed to recognize these kinds of characters. However, since no definition was made for documents, characters without noise were studied in practice. Bilim Kodu Anahtar Kelimeler Sayfa Adedi Tez Yöneticisi 646.04.01 Character recognition, skeletion 121 Doç.Dr.F.Nihal GÜLER