Some studies on text detection in Borndigital and scene text images

Abstract

Images are the most valuable and easily available media and a great newlineresource of information Text information is influential as it plays a major role newlinein our day to day life. Text provides useful semantic description about the newlinecontents of an image or video which is direct high level semantic newlineinformation The process of extracting regions of the image containing text is newlineknown as text detection Identifying text regions in an image is a challenging newlinetask because there can be many variations in the image due to uneven newlinebackgrounds variations in contrast text size style color and so on Also newlinethere are problems of false positives That is text regions being considered as newlinenon text regions can affect the semantics of the text To overcome these there newlineis a need for better text detection algorithms and text features that can be good newlinedescriptors of text regions. newlineNew text detection methods have been developed for efficiently newlineextracting text information from images. The first work uses feature based newlinetetra map to detect text in born digital images newline newline