Hybrid OCR Techniques for Cursive Script Languages - A Review and Applications

Software-based Arabic optical character recognition (OCR) has been used quite successfully for many years. However, the hardware-based implementations of the OCR - which can be 10-100 times faster than the software-only method - seem to not have been fully exploited. This paper briefly reviews the research material addressing software based OCR but focuses more on the hardware realization of Arabic OCR. The software-only OCR methods have achieved reasonable maturity but require use of PCs (or similar platforms), which is an obvious hindrance for OCR implementations in small form factor, such as pens or mobile devices. Different researchers have addressed individual, sub-tasks of Arabic (non-Latin) OCR, but not as complete, functional systems. The recognition accuracies also have significant room of improvement. The main challenge for practically functioning OCR systems in smaller sizes is the optimization of software algorithms so they can be efficiently and cost-effectively realized in hardware.