Automatic text localization in natural scene images

Abstract

In this paper we present an approach to automatic localization of text in natural scene images. Text which is embedded in a natural scene, e.g. in the street, shop or a bus stop, is not available to visually impaired persons, therefore it is necessary to design systems for automatic localization and recognition of such text in a image of such a scene. We propose an approach that utilizes a novel corner measure and is based on the assumption that areas of image containing text exhibit a large density of edges forming corners. We show that the proposed method combines computational simplicity and very good precision and recall on a well-known reference image database. © 2011 Springer-Verlag Berlin Heidelberg.