Research On The Key Technology Of Improving The Quality Of Multiple Social Videos

Recently, because of the development of mobile phones and other portable camera equipment, massive videos and films which shot by amateurs emerged. Their popularity has a great impact on people’s lives. Nonprofessional videos often have severe jitters, and can never be appreciated normally. So it is important to increase the video quality by means of some methods.In this paper, a new algorithm is introduced to increase the quality of the multiple perspective videos and the egocentric videos. It has three parts, the egocentric videos stabilization, prediction of joint attention based on multiple perspective videos and multiple egocentric videos automatic editing.The main work of this paper are:(1) An egocentric videos stabilization algorithm is proposed. Firstly, it estimates the positions and orientations of the camera. Secondly, a virtual path of the 3D motion of the camera is obtained by polynomial curve fitting. Then, the orientations is smoothed by the vector interpolation. Finally, the video frames is selected and modified based on the virtual path curve and the smoothing orientations. Experimental results on real data sets show that the proposed algorithm is effective for the videos of portable cameras.(2) A joint attention estimating algorithm based on multiple perspective videos is proposed. Modeled the observer’s joint attention in the scene and defined the distance between observers and their joint attention. The iterative formula of mean shift algorithm of multidimensional kernel density estimation is derived. And the experiment on real data sets demonstrates the effectiveness of the proposed algorithm on multiple perspective videos.(3) A multi view videos editing algorithm is introduced. It appraises the video clips by the joint attentions, the time and the distance between the observers and the objects, and finally models videos editing process by the directed graph to look for the best way to edit a fine video. The experimental results show that the proposed method is effective.