An Approach for Video-Text Extraction Based on Text Traversing Line and Stroke Connectivity

In complex background, conventional automatic video-text location methods can not robustly locate text. A robust video-text location method is proposed in this paper. It can be divided into two stages. In the first stage, an unsupervised paradigm based on wavelet is applied to obtain candidate text region. In the second stage, traversing line with its aptitude spectrum is introduced and applied to create boundary for each text line. Experiment demonstrates that our scheme is robust to text of various color, font, sizes in complex image.