Farsi Word Spotting and Font Size Recognition

This paper is the first paper about Farsi word spotting and font size recognition. In this work using some font size independent features such as the number, aspect ratio, and mesh features of sub words, a Farsi keyword is described, searched and found. Also font size of document image is recognized. This approach has been evaluated on a dataset consisting of 500 Farsi document images where font size recognition rate of 94.2%. Retrieval precision rate of 92.3% at recall rate of 76.5% has been obtained. This approach with little adaptation is applicable on Arabic and Urdu documents.