Research On Efficient Method For Video Stabilization And Content Exploration

In the video capture process, because of the limitations of equipment and environmental, photographer can not keep camera stable for a long time, resulting in jitter effects existint in the video, affecting viewing experience and post processing. When gathering huge amounts of video data, users need to see every frame in the video content browsing and searching, with time-consuming process and shitty user experience, making against to achieve video information efficiently.With regard to existing problem in video stabilization, this paper presents a video stabilization method based on global optimization, simultaneously solving motion compensation and image warping.Using video feature point trajectories modeling camera motion, this paper uses similarity invariant to convert motion trajectories to representation of mesh vertices, so we can obtain image warps parameters of video stabilization by solving energy minimization problem. Experimental results show that this method can reduce frame image distortion after stabilization, while assuring the efficiency of video stabilization.With regard to the efficiency of video content browsing, this paper presents an efficient two-dimensional graph structure based video browsing method. Unlike the traditional linear video browsing method based on progress bar, this paper visualizes video scene information by scenegraph, and provides more efficient nolinear video browsing method for users. By computing similarity of video shots, scene structure can be derived by clustering video shots. Using automatic drawing technique of graph, scenegraph is reconstructed to present video scene topology structure, with node representing scene and edge representing scene transition. Experimental results show this method can guide users to browse more video data efficiently in limited time.