No abstract

This paper presents an approach to movie video indexing that utilises audiovisual analysis to detect important and meaningful temporal video segments, that we term events. We consider three event classes, corresponding to dialogues, action sequences, and montages, where the latter also includes musical sequences. These three event classes are intuitive for a viewer to understand and recognise whilst accounting for over 90% of the content of most movies. To detect events we leverage traditional filmmaking principles and map these to a set of computable low-level audiovisual features. Finite state machines (FSMs) are used to detect when temporal sequences of specific features occur. A set of heuristics, again inspired by filmmaking conventions, are then applied to the output of multiple FSMs to detect the required events. A movie search system, named MovieBrowser, built upon this approach is also described. The overall approach is evaluated against a ground truth of over twenty-three hours of movie content drawn from various genres and consistently obtains high precision and recall for all event classes. A user experiment designed to evaluate the usefulness of an event-based structure for both searching and browsing movie archives is also described and the results indicate the usefulness of the proposed approach.

Movie trailers, or previews, are an important method of advertising movies. They are extensively shown before movies in cinemas, as well as on television and increasingly, over the Internet. Making a trailer is a creative process, in which a number of shots from a movie are selected in order to entice a viewer in to paying to see the full movie. Thus, the creation of these trailers is an integral part in the promotion of a movie. Action movies in particular rely on trailers as a form of advertising as it is possible to show short, exciting portions of an action movie, which are likely to appeal to the target audience. This paper presents an approach which automatically selects shots from action movies in order to assist in the creation of trailers. A set of audiovisual features are extracted that aim to model the characteristics of shots typically present in trailers, and a support vector machine is utilised in order to select the relevant shots. The approach taken is not particularly novel but the results show that the process may be used in order to ease the trailer creation process or to facilitate the creation of variable length, or personalised trailers.

Abstract. The recent past has seen a proliferation in the amount of digital video content being created and consumed. This is perhaps being driven by the increase in audiovisual quality, as well as the ease with which production, reproduction and consumption is now possible. The widespread use of digital video, as opposed its analogue counterpart, has opened up a plethora of previously impossible applications. This paper builds upon previous work that analysed digital video, namely movies, in order to facilitate presentation in an easily navigable manner. A film browsing interface, termed the MovieBrowser, is described, which allows users to easily locate specific portions of movies, as well as to obtain an understanding of the filming being perused. A number of experiments which assess the system's performance are also presented.

This paper presents an approach to movie video indexing that utilises audiovisual analysis to detect important and meaningful temporal video segments, that we term events. We consider three event classes, corresponding to dialogues, action sequences, and montages, where the latter also includes musical sequences. These three event classes are intuitive for a viewer to understand and recognise whilst accounting for over 90% of the content of most movies. To detect events we leverage traditional filmmaking principles and map these to a set of computable low-level audiovisual features. Finite state machines (FSMs) are used to detect when temporal sequences of specific features occur. A set of heuristics, again inspired by filmmaking conventions, are then applied to the output of multiple FSMs to detect the required events. A movie search system, named MovieBrowser, built upon this approach is also described. The overall approach is evaluated against a ground truth of over twenty-three hours of movie content drawn from various genres and consistently obtains high precision and recall for all event classes. A user experiment designed to evaluate the usefulness of an event-based structure for both searching and browsing movie archives is also described and the results indicate the usefulness of the proposed approach.