An Overview+Detail Surveillance Video Player: Information-Based Adaptive Fast-Forward

In this paper, we propose an adaptive fast playback framework where multi-features are used to support arbitrary frame-rate video playback. We introduce a Jenson noise-based difference (JSND) as a distance measure between adjacent frames for video key frame extraction, and then present an interest learning model to control the playback rate according to the user preference. The proposed “smart-skip” frame schema not only helps users navigate the video content non-uniformly for any variable playback rate, but also preserves video semantic information to avoid the omission of important events. An overview+detail video player offering users an immersive experience is implemented to browse and comprehend the video content. Experimental results show that users can quickly skim the video, understand the content, and navigate into the content of interest.