Video Key Frame Extraction Using Tsallis Entropy And Jensen Divergence

With the coming of the Internet era,mobile intelligent terminal equipment can be seen everywhere,which video capture function has been widely used,more and more video information appears in our life.The diversity of video types,the richness of video content,all give us a big problem in understanding.Whether we can understand the video content fast and effectively in a short time becomes crucial.However,to analysis and understand means to deal with a large amount of data,which is not a simple work in practical application.Video key frame extraction is an effective solution to the above requirements,through a set of streamlined,representative video frame sequence-video key frame,to express the main content of the original video.In this thesis,the relevant knowledge of key frame extraction is summarized.Based on this,a new method for video key frame extraction is proposed.This method obtains the difference values between continuous video frames through the Jensen-Tsallis divergence,and according those outliers to locate the shot boundary,so that the video segment is segmented into multiple lenses.In order to refine the complexity of the shot content,we divide the different types of sub shots according to the intensity of video content.In addition,this thesis analyzes the selection of the optimal Tsallis entropy index from two aspects,and it is believed that the appropriate entropy index can better reflect the advantages of Tsallis entropy in evaluating the probability distribution.Through the experimental tests of various types of videos,the results show that the proposed method is fast and efficient.And the results are better than the reference method in both objective index calculation and subjective evaluation,so,the method in this thesis can achieve the real-time and universal effect basically.