A segmentation-free approach to recognise printed Sinhala script using linear symmetry

In this paper, a novel approach for printed character recognition using linear symmetry is proposed. When the conventional character recognition methods such as the arti1cial neural network based techniques are used to recognise Brahmi Sinhala script, segmentation of modi1ed characters into modi1er symbols and basic characters is a necessity but a complex issue. The large size of the character set makes the whole recognition process even more complex. In contrast, in the proposed method, the orientation features are e7ectively used to recognise characters directly using a standard alphabet as the basis without the need for segmentation into basic components. The edge detection algorithm using linear symmetry recognises vertical modi1ers. The linear symmetry principle is also used to determine the skew angle. Experiments with the aim for an optical character recognition system for the printed Sinhala script show favourable results. ? 2004 Pattern Recognition Society. Published by Elsevier Ltd. All rights reserved.