A Performance Comparison of Segmentation Techniques for the Urdu Text

Segmentation is a procedure of splitting the image contents into its subparts (i.e., line and words). For all the common language-handling applications, for example document structure extraction, content rebuilding, optical character recognition, falsifications, security, graphology, and so forth, segmentation is an indispensable and primary step. This paper presents the quantitative performance of three different text line segmentation techniques: projection method, smearing method, and edge information-based method, for the Urdu Nastaleeq type-written text. The evaluation is performed over the gathered standard data samples taken from different magazines, poetry books, and newspapers, using precision and recall metrics. In the course of evaluation, the potency and debility of algorithms are analyzed and it is spotted that smearing segmentation method checkmates the other two methods.