Automatic caption extraction of news video and its implementation

Text information retrieval is an important field in content-based retrieval of video. In this paper a new advanced algorithm is presented. This algorithm detects video shots and text events firstly, and then locates the caption region with the adaptive method and projection method respectively. In order to get enhanced character, frame averaging method is employed in a series of video frames with captions. Finally, the obtained colorful character area is segmented with Fuzzy c-means method. Experimental results show that this caption detection scheme is effective and robust.