A novel free format Persian/Arabic handwritten zip code recognition system

Article history: Received 13 January 2012 Received in revised form 22 April 2013 Accepted 22 April 2013 Available online xxxx In Iran like many other countries, the categorization of postal envelopes is executed manually, mostly based on the handwritten addresses and zip codes. That process is still slow and prone to man-made errors. Therefore, having an automated, accurate and efficient system to recognize handwritten zip codes is of high necessity for a faster and easier arrangement of postal envelopes, and consequently, enhanced performance of the post office. A complete system for Persian (Farsi) handwritten zip code detection and recognition is introduced in this paper. The proposed system consists of two phases; with zip code localization being the first and the main contribution of the study, and zip code digits recognition as the second. The first phase, proposes a state-of-the-art algorithm which localizes the zip code on the envelope. Subsequent to digit segmentation, handwriting zip code digits are recognized in the second phase, which includes feature extraction and classification sub-steps. The results obtained from the 50 test samples show an overall recognition rate of 92.9% in localization and recognition of handwritten zip codes. 2013 Elsevier Ltd. All rights reserved.