Phase congruency and morphology based approach for text localization in videos

摘要

In this knowledge era, cognitive learning and media technologies promote high visual orientation through videos which appears in the form of text, graphics, animations, audio or still images. Text present in videos carry important semantic information which is indeed essential for video comprehension. In this context, we propose a novel method of detecting the text clusters in video frames based on the phase congruency model and morphology based approaches. We investigate the matching features extracted from both these methods and devise a set of rules using morphological operators for false positive elimination. The text regions are detected using connected component analysis and finally the text is localized. We have evaluated the performance of the proposed method on the standard datasets and the results highlight the effectiveness in localizing the text in videos and scene images.