Text detection in video images using adaptive edge detection and Stroke Width verification

Text displayed in a video provides important information about the video content. Therefore, it can be utilized as a valuable source for indexing and retrieval in digital video libraries. In this paper, we propose a novel approach for efficient automated text detection in video data: Firstly, we developed an edge-based multi-scale text detector to identify potential text candidates with high recall rate and small computational time expenses. Secondly, candidate text lines are refined by an image entropy based improvement algorithm and a Stroke Width Transform (SWT) based verification procedure. Both types of text, overlay and recorded scene text can be localized reliably. The accuracy of the proposed approach is proven by evaluation.