Abstract
In this paper we present a text detection and localization method Our detection technique is based on a cascade of boosted ensemble and localizer uses standard image processing techniques. We propose a small set of features (39 in total) capable of detecting various type of text in grey level natural scene images. Two weak learners, linear discriminant function and log likelihood-ratio test under gaussian assumption, are evaluated. Single features and combination of features are used to form weak classifiers. The proposed scheme is evaluated on ICDAR 2003 robust reading and text locating database. The results are encouraging and the detector can process an images of 640x480 pixels in less than 2 seconds.