Conclusion and Future Work
This book presents substantial contribution in the field of document image analysis specifically in cursive scene text recognition. The complexities of Arabic script have been highlighted, and potential of RNN is exploited because of its context learning ability. The RNN-based statistical models were adapted to support the learning of complex Arabic script. In cursive scripts, there are numerous works presented for Chinese and Japanese characters but Arabic script has not exposed to state-of-the-art research methods. Few efforts are reported to date, which obviously are not enough to address the complexities. The scene text recognition can be categorized as a subproblem of Optical Character Recognition (OCR). Although the appearance of samples do not share similarities. Likewise, there is much difference between handwritten and printed Arabic text. The printed text is more clean and clear whereas handwritten text is dominated by writing styles along implicit noise attached with scanned document.