Research On Video Abstraction Techniques

As the growing of internet and multi-media technology, digital video booms, such as news, advertisement, movies, home video and etc. All of these video applications are facing the emerging digitalized video data in a large amount. But as we know, video information is linear and non-structural, which makes video browsing to be time-consuming and boring. Video abstraction, aiming at reducing redundant data and improving browsing efficiency, is very important in various video-based applications. Video abstraction system has been one of the research focuses in recent years.In this paper, we first review some key technologies of video abstraction. And then we study a method based HSV-color space model of scene cutting in video abstraction. In the method, the first step is to transfer the RGB values of the picture frame to the corresponding HSV values. After doing non-equal interval quantizing to the three HSV weights, the next is to compound the quantizing weights to one-dimensional eigenvector. Calculating the one-dimensional eigenvector histograms of the adjacent two frames can get the similarity. According to the similarity we can judge if the video scene has converted.The follow we research a method based two times clustering of key frame extracting in video abstraction. First we calculate the difference characteristic among the adjacent two frames based on segment weighted color histogram, than all the frames are clustered to class using adaptive threshold algorithm, the nearest frame and farthest frame to the center of its class is automatically extracted as two key frame in the shot. Experimental results demonstrate that the method can better distill the key frame in the video order.At the end of this paper, we give a system implementation for video abstraction of digital film. It is an experiment to put some of the popular video analysis technology into practice.