Comparative study on recent text line segmentation methods of unconstrained handwritten scripts
Ayush Pradhan, Sidharth Behera, Pushpalata Pujari · 2017
Different National archives and libraries across the world have preserved available historical handwritten documents which needs electronic transformation for automatic recognition. For achieving this, the line segmentation is a decisive preprocessing stage for document structure extraction and character recognition. If the quality of these documents is poor due to background noise, artefacts and interfering lines, automatic segmentation of text lines is still a burning research issue amongst the natural language processing (NLP) community. During the decade, 1993-2005, a significant progress in this field has been made, but many technical issues still remained unsolved. In the meantime further one decade has passed and therefore there is curiosity to explore the latest status on handwritten line segmentation technology or to find any improvement has been made on any of the existing methods. Keeping this objective in view the present survey has been carried out and an overview of new and modified methods on text line segmentation of handwritten archives published during 2006-2015 are presented. In addition, concise presentation of the methods published during 1993-2005 is also included so that it could benefit to the researchers who are either new or currently carrying out research work in this field.