Visual Analytic-Based Technique for Handwritten Indic Script Identification—A Greedy Heuristic Feature Fusion Framework

摘要

Script identification from multi-script handwritten document images has been a subject of considerable discussion in the literature. In this paper, a novel feature fusion framework (FFF) using structural appearance (SA) and directional morphological filter (DMF) is proposed based on the idea of visual analytic (VA). A dataset of 181 handwritten document pages distributed over 2450 line and 20,260 word images of Bangla, Devanagari, Roman, Oriya, and Urdu scripts is built and considered for experimentation. Experimental result shows a significant improvement of the identification rate by the VA-FFF over the SA and DMF technique if they had been applied individually.