Unsupervised Video Summarization via Dynamic Modeling-Based Hierarchical Clustering

Abstract

Mining the video data using unsupervised learning techniques can reveal important information regarding the internal visual content of large video databases. One of these information is the video summary which is a sequence of still pictures that represent the content of a video in such a way that the respective target group is rapidly provided with concise information about the content, while the essential message of the original video is preserved. In this paper, an enhanced method for generating static video summaries is presented. This method utilizes a modified dynamic modeling-based hierarchical clustering algorithm that depends on the temporal order and sequential nature of the video to fasten the clustering process. Video summaries generated by our method are compared with summaries generated by others found in the literature and the ground truth summaries. Experimental results indicate that the video summaries generated by the proposed method have a higher quality than others.