Text in images and videos is vital for understanding the visual content. In this paper, we propose to combine different features (namely corner, stroke width similarity and color similarity) to detect Chinese text in complex images and videos. The corners are used to determine the potential text candidates which are then refined using stroke width and color features. To further enhance the efficiency of the detection algorithm, a line scanning strategy is adopted to select the correct text regions. A new challenging data set with ground truth and evaluation protocol is built and will be made publicly available for research purposes. It is collected from different TV programs. Extensive experimental analysis shows that our proposed algorithm yields in very promising results which compare favorably against traditional approaches in the research literature.