| Realizing intelligent analysis and understanding of the digital video has become a hot spot in the field of information analysis.As the basic processing algorithms,video shot segmentation technology has been widely studied.However,the simple shot segmentation does not reflect the video content from the semantic planes,especially for the specific types of video analysis,such as news video.In order to meet the above requirements,the concept of news basic unit was proposed in this paper.Then a news story segmentation algorithm based boundary classification and a news abstract extraction algorithm based content analysis were presented.A new boundary classification based news story segmentation algorithm was proposed.Firstly,news basic unit was defined.A news basic unit is a video clip which begins with a section of the anchor shot,closely follows with content shot and regards the start moment of next anchor shot as end.By using news basic unit,this paper transfers the news story segmentation into boundary classification,which can improve segmentation precision effectively.The news basic unit can reduce the semantic gap between the video shot and news story,which is conducive to the news semantic analysis.There are two processes of proposed news segmentation algorithm: extracting and analyzing the news basic unit.With the help of news basic unit,video structure analysis can be achieved and the story segmentation problem can be translated into boundary classification problem.Firstly,shot segmentation based histogram and shot demarcation based anchor template match were used to gain news basic unit.Then,audio and video features were analyzed to describe news basic unit and gain the final segmentation result.The experimental results show that the proposed scheme can express story boundaries effectively and divide the news into story unit precisely.A new content based news abstract extraction algorithm proposed in this paper.This paper synthesized video features and topic caption information to develop a news static abstraction algorithm,which can select key frames adaptively to make up abstraction.This algorithm can remove the redundancy and summarize news semantic.First of all,two principals were put forward in this paper.Then according to the principal,an efficient abstract extraction algorithm was designed to summarize the content of the news.The experimental results show that the proposed algorithm can remove the redundant of news and reserve the significate content.Finally,this paper simulated a news video analysis platform.The above two algorithms were integrated to develop a content based news video analysis platform.This platform was developed by C++ and Qt in Linux,could segment news story accurately and summarize news content effectively,which provides convenience for subsequent news analysis. |
PDF Full Text Request