Yadong Wu, Song Wang, Shuiqiang Lin, Yongguo Han
In order to clearly visualize the video structure and provide a user-friendly video content review and retrieve system, we firstly apply total variation (TV) metric to video shot segmentation and propose a novel video visualization method in this paper. Firstly, video data visualization design principles are proposed based on the analysis of video data characteristics. Secondly, a TV metric is proposed to measure the distance between two video frames. Then, video shots are detected by TV metric, and video key frames are extracted by K-means algorithm. Finally, according to the proposed video visualization design principles, a spiral video visualization method is proposed. Experimental results show that the proposed video visualization method can express the video content and the whole structure directly and clearly.