Reconstruct Damaged Sinhala Handwritten Characters by Guided Edge Linking Approach

The Sinhala language is considered a low-resource language in the computing domain. Before reconstructing damaged characters, damaged areas should be identified. The fundamental process in the damaged Sinhala handwritten character reconstruction process is to identify features of the damaged end tip pixels and character segments. Discovering suitable end tip pixel couples and linking them via appropriate linking shapes (straight or curved) is di cult because most Sinhala characters consist of curved shapes compared to other language characters like English. The most suitable end tips are discovered by considering both the minimum distance end tips and the direction of each end tip in the damaged Sinhala handwritten character. Edges of the damaged Sinhala handwritten characters are linked through local and global edge connecting using straight-line end tip linking and the Bezier curve end tip linking approaches, respectively. The results show that the proposed direction-based nearest end- tip coupling approach provides 96.2% accuracy, while coupled end-tip linking approach gives an accuracy of 94.2%.