Abstract
It is hard to achieve the goal that making computer understand a film content, because there are too many events may happen in a film. Since the difficult the film content in computer understanding, we first experiment how the low level visual features can affect a film. Based on some low level visual features have been proposed, we presents a framework for the classification of feature films into genres in this thesis. Our current domain of study is the movie preview. A film preview often emphasizes the theme of a film and hence provides suitable information for classification, on the other hand, a preview is more easy to get than a whole movie. In our approach, we classify movies into three broad categories: Action, Dramas,or thriller films. In our experiment, four computable video features (average shot length, color variance, motion content and lighting key) are analysis that how well these features for classify films into genres. On the other hand, we proposed another features called visual effect to distinguish drama or non-drama films. Classification Tree and Neural network are used after analysis of these low level visual features, with the features distribution. After our experiment, we found that our approach can also be broadened for other potential applications includingthe building and updating of video databases with minimal human intervention, scene understanding, browsing and retrieval of videos on the Internet and video libraries.