Visual Saliency and Boosting Based Background Suppression for Scene Text

To solve the issue of background interferences on the scene text automatic localization algorithm, a scheme of background suppression for scene text is proposed, which utilizes characteristic of visual saliency to combine histogram of oriented gradient features, its statistical features, gradient magnitude features and gradient curve features with the boosting frame. The scheme aims to suppressing the complex background and highlighting the foreground text in natural scene. It can consider be as the preprocessing stage of the scene text automatic localization algorithm, and it improves the performances of the scene text automatic localization algorithm. The experimental results in both the ICDAR2011 scene text localization competition test dataset and the laboratory Chinese dataset show that the proposed scheme can suppress effectively the complex background and improve the scene text localization algorithm.