Key-frame extraction for video captured by smart phones

In this paper, the key-frame extraction for video clips which are captured by smart phones are investigated. A weighted temporal consistent fuzzy clustering technology is used to cluster the video frames. The image-level feature and the orientation information are combined into the feature vector and the accelerate values are used to determine the weight. Such a method sufficiently exploits the spatio-temporal characterizes of the video clips captured by smart phones. The algorithm is implemented in a Sumsung Nexus S smart phone and we make an extensive experimental validation.