Tackling action-based video abstraction of animated movies for video browsing

In this paper we address the issue of producing automatic video abstracts in the video indexing context of animated movies. For a quick browse of the movie’s visual content, we create first a storyboard-like summary, which follows the movie events by providing one key frame for each particular scene. To capture the shot visual activity, we use histograms of cumulative inter-frame distances and key frames are adapted to histogram mode distribution. Secondly, for a ”sneak peak” of the movie’s exciting action content, we propose a trailer-like video skim, which aims at providing only the most interesting parts of the movie. Our method is based on a relatively standard approach, thus highlighting action through the analysis of the movie rhythm and visual activity information. To fit every movie content, including predominant static movies or movies without exciting parts, the concept of action gets tuned according to movie average rhythm. The efficiency of our approach 1 has been confirmed through several user studies, directly by the ”consumers of the product”, i.e. the end-users.