| With the development of the intelligent equipment and the Internet,it becomes more and more convenient for people to capture videos.It is a big challenge to get the useful information from the massive video data.Video summarization is a summary based on the video visual content,this summary is different from using text to describe a video,but generate dynamic video summary which we could watch.And this summary can not only retain the useful information of the original video,but also enables to make a good summary of the video content.This thesis researches the most popular video summarization algorithms,analyzes the key technologies used in the generation of video summarization.Owing to the existence of many problems in the field of video summarization,we proposed using unsupervised learning method for the research of video summarization.There are two big advantages about the proposed method:First,the proposed method enables to generate video summarization without training.Second,the proposed method could save our time because we can get video summarization in a relatively short period time by using this method.The main work of this article is as follow:(1)We proposed a dynamic video summarization method with the fusion of multiple features and image quality.This method using multiple features to represent video,and fusing the image quality to evaluate the importance of the video clip,finally using this importance to generate high-quality video summary.(2)We proposed a dynamic video summarization method by fusing the global importance and the local importance.Not only the global importance is considered for the generation of video summary,but also the local importance is considered.Finally using the fusing importance to generate high-quality video summary.(3)We design and implement a video summarization system.This video summarization system can quickly process videos and generate good video summaries by feeding with users'video and setting up the length of video summary. |
PDF Full Text Request