A novel method to reduce redundancy in adaptive threshold clustering key frame extraction systems

In adaptive threshold clustering key frame extraction systems, the video is sliced into different segments according to the adaptive threshold. One or more frames are selected from each segment as key fames to represent the video. However, those key frames may very similar. These redundant key frames provide limited information. In this paper, we propose a novel method to handle this problem which removes redundancy based on the edge structure similar measure. Experiment results show that this method is effective in reducing the redundancy comparing with methods based on low-level color information in term of accuracy and the time complexity.