Analysis of Line Segmentation Methods for Historical Manuscripts

Historical manuscripts are rich source of knowledge about culture, civilization, political system, ancient science and mathematical research and also represent the cultural as well as the religious heritage of the country. Handwritten Text Recognition is the technique to convert the handwritten document into a machine-understandable format. Line segmentation of the document is a very crucial step of the HTR. Last few decades, researchers have invented many algorithms for the segmentation of handwritten or printed documents but these algorithms are not directly applied on these manuscripts due to the presence of the touching lines, uneven space between lines, touching characters, slant, stains, different font size and writing styles, decoration, noise, wrinkles, faded ink, ink bleed through, stitches, rust etc. These issues make the segment process bit complicated. In this paper, an attempt is made to report various traditional and learning based line segmentation methods along with their performances.