An efficient scheme for tilt correction in Arabic OCR system

Preprocessing stage is required in almost every image processing application ranging from biometric analysis to document image analysis. An input image or information need to be normalized and converted into format acceptable by OCR (optical character recognition) system. OCR systems typically assume that documents were printed with a single direction of the text and that the acquisition process did not introduce a relevant skew. Practically this assumption is not very strong and printed documents could be skewed at some angle with horizontal axis. In this paper, we have proposed skew estimation of document images for Arabic fonts. It is based upon the specific feature of Arabic script. In our proposed scheme, we scan for the occurrence of letter ´alif´ and estimate the tilt based upon its slope. Extensive experimentation was performed and scheme was found to be very effective.