An Approach for Resolving Double Character Segmentation in Sinhala Social Media Text Images

摘要

Sinhala is the official language of Sri Lanka, which has descended from the ancient Brahmi script. For Asian languages such as Sinhala and Tamil, only a little attention has been given to text extraction. Sinhala text in digital images have different fonts and font sizes, thus making the text extraction process harder. It is a challenging task to segment overlapping and touching Sinhala characters in digital images, and effectively recognize the segmented characters. This paper outlines a software system designed and implemented to extract Sinhala text from digital images. In this study, a connected components labelling algorithm was used for overlapping character segmentation, while a background thinning based approach was used for touching character segmentation. The proposed system is capable of extracting Sinhala text from digital images with nearly 76% accuracy.