Segmentation of news items from newspaper images

Abstract

Segmenting news items from newspaper images is a challenging task due to the diverse layouts and overlapping content. In this paper, we propose an algorithm that utilizes the lines separating articles as key features for accurate segmentation. The algorithm focuses on identifying the rectangle points formed by vertical and horizontal lines. To provide a user-friendly interface, we develop a Web-API using Flask. Users can easily upload newspaper images and obtain segmented news items as output. The experimental results demonstrate the effectiveness of our approach in accurately segmenting news items from newspaper images.The algorithm is tested on a set of complex English newspaper images such as The Hindu, The Telegraph and The Statement.