A complete framework for video temporal segmentation

In this paper we propose a complete high level segmentation algorithm of video flows into scenes. In the first phase we identify shot boundaries with an accuracy of more than 95% in recall and 90% in precision rates at reduced computational time. In a second stage, a storyboard is created by using a leap keyframe extraction method, at 27% faster rate than the reference method for equivalent performances. Finally, the detected keyframes feed a shot clustering algorithm which integrates a set of temporal constraints that generates video scenes with an average precision and recall rates of 73.7% and 84.6% respectively.