Automatic Chapter Generation for Consumer Videos

This paper proposes a novel framework for automatic chapter generation of consumer videos, in which the knowledge of media aesthetics and video production is utilized. For an input video, a set of specific visual and aural aesthetic features, called expressive elements, is first extracted. Then a high-level structure with aesthetic semantics, namely aesthetic intensity, is obtained and used to generate the corresponding video chapters. Experimental results demonstrate the effectiveness and efficiency of the proposed approach.