Morphological gradient based approach for text localization in video/scene images

In this work, we present an approach for detecting the text present in videos/scene images based on morphological gradient information. The system detects the gradient information using morphological operations and the obtained results are binarized. The resultant binarized image contains some non-text regions which are then morphologically opened so that the small components with less than 4-pixel connectivity are eliminated producing another binary image. Finally, we employ connected component analysis and morphological dilation operation to determine the text regions and hence to localize text blocks. The experimental results obtained on publicly available standard datasets illustrate that the proposed method accurately detect and localize texts of various sizes, fonts and colors in videos and scene images.