Accurate graph-based scene segmentation using object matching and audio feature

A method for accurate scene segmentation utilizing two kinds of directed graph obtained by object matching and by using audio features is proposed. Generally, in audio-visual materials, there are repeated appearances of shots that include frames of the same background, object or place. It should be assumed that such shots are included in a single scene. In the proposed method, by performing object matching between two frames that are each included in different shots, multiple shots for which frames include the same object can be successfully found and their repeated appearances are represented as a directed graph. The proposed method also generates another directed graph that represents the repeated appearances of shots with similar audio features. By the combined use of these two graphs, accurate scene segmentation can be realized. The effectiveness of the proposed method is verified by applying this method to news programs and another broadcast program.