An optional character recognition system for Maltese

Abstract

Optical Character Recognition (OCR) entails the process of converting scanned images of machine printed or handwritten text (numerals, letters, and symbols), into a computer processable format. The aim of this project is to develop Character Recognition for Maltese text since OCR systems developed for the English language, or other languages, cannot be directly used for Maltese, as the latter contains some specific characters that are invariably misinterpreted. The proposed method, based on a feature extraction approach, extracts structural and positional primitive features for each symbol. The number of features extracted for each symbol is maintained fixed. In this way the tested method can be plugged as an input to a Neural Net to improve performance. A complete recognition system was implemented able pre-process an image, recognize and export the result to existing applications. Evaluation and testing of the complete recognition system show an average 90 percent recognition rate.