Research On Video Summarization Based On Peak Density Clustering And Network Community Structure

Digital video is an important channel of multimedia technology and is widely applied in all aspects of life in society.How to enable users to quickly capture the content of the video and decide whether to continue watching is a problem that needs attention.In this context,video summarization technology came into being.Video summarization is a new content-based video compression technology,which can effectively find important information from the video and eliminate redundant data.In recent years,video summarization technology has made great progress,but how to generate high-efficiency and high-accuracy video summaries is still under constant exploration.The research of video summarization technology is studied in this thesis.The main works are as follows:(1)A video summarization algorithm based on density peak clustering(DPC)is proposed.First,the features of the hue histogram of the video frame is extracted and is clustered using DPC.The video shot is segmented according the clustering results.Secondly,in the keyframe extraction stage,select the frame closest to the cluster center as the keyframe.At the same time,the histogram intersection method is employed to remove the similar frames in the keyframes to generate a video summary.Experiments show that this method has obtained better video summary results.(2)A video summarization algorithm based on the density peak clustering algorithm with temporal characteristics(T-DPC algorithm)is presented.The time characteristics of video frames are added to the DPC algorithm,and a density peak clustering algorithm with time characteristics is proposed and applied to shot segmentation.In the keyframe extraction stage,the information entropy of the video frame and the degree of similarity to the cluster center are taken into consideration to extract keyframes,and similar frames are eliminated.The experimental results show that the quality of the video summary obtained by T-DPC method is higher than that of the comparison algorithms.(3)A video summarization algorithm based on Fast Newman community detection algorithm,called(FN algorithm)is proposed.FN algorithm regards video frames as network nodes,and uses the similarity between frames as the weight of the edges to construct a fully connected undirected weights network.The Fast Newman algorithm is applied to the network to detect the community,and the frame closest to the average degree of node is extracted in each community as the keyframe.At the same time,the color feature and the SURF feature are used to remove similar frames,and an ideal video summary is obtained.