Arabic Font Recognition Based on Templates
Department of Electrical and Computer Engineering, Islamic University of Gaza, Palestine
Abstract: We present an algorithm for a priori Arabic optical Font Recognition (AFR). First, words in the training set of documents for each font are segmented into symbols that are rescaled. Next, templates are constructed, where every new training symbol that is not similar to existing templates is a new template. Templates are sharable between fonts. To classify the font of a word, its symbols are matched to the templates and the fonts of the best matching templates are retained. The most frequent font is the word font.
Keywords: Optical character recognition, optical font recognition, vertical normalization, template matching.
Received January 29, 2003; accepted May 4, 2003