Video scene segmentation using time constraint dominant-set clustering
Video scene segmentation plays an important role in video structure analysis. In this paper, we propose a time constraint dominant-set clustering algorithm for shot grouping and scene segmentation, in which the similarity between shots is based on autocorrelogram feature, motion feature and time constraint. Therefore, the visual evidence and time constraint contained in the video content are effectively incorporated into a unified clustering framework. Moreover, the number of clusters in our algorithm does not need to be predefined and thus it provides an automatic framework for scene segmentation. Compared with normalized cut clustering based scene segmentation, our algorithm can achieve more accurate results and requires less computing resources.