Video Summarization Based On Hypergraph Ranking

With the rapid development of multimedia and Internet technology,Video summarization has received widely attention as a technique to quickly display the main video content.Existing graph model based method takes video frames as vertices and uses the edges to build the relationship between two vertices,although good results are obtained,which may not well capture the complex relationship among the video frames.To overcome this drawback,this paper employs the hypergraph model to capture the complex and higher-order relations in a video,and make a deep research and analysis on the application of hypergraph model in video summarization field.Firstly,we propose a new method based on hypergraph ranking to generate a static video summarization,and name it HGRVS(Hypergraph Ranking based Video Summarization).The HGRVS method includes four steps,video feature extraction,video hypergraph construction,video frames classification and video summarization generation.The main idea of HGRVS is to classify video frames by using hypergraph ranking and selecte candidate keyframes from each frame classifications,then an objective function is proposed which combined with maximize unorrelation,maximize information coverage and minimum information redundancy to determine the finall video summarization.Secondly,RWH is proposed which based on random walk on Hypergraph.This algorithm regards the generation of video summarization as a sorting problem.By random walk,we can compute the stationary distribution probability of each frame,and video frames can be sorted.The larger probability of frame,the more important it is.Then summary is produced by eliminating similar keyframes from frames with larger probability.Finally,extensive subjective and objective experiments clearly demonstrate the superiority of HGRVS and RWH to the state-of-the-art approaches.