Edge Restoration of Video Stabilization Based on Panoramic Stitching

Abstract

Video stabilization is widely used in various fields of life. After video stabilization, due to the offset of video frames, it often brings the problem of edge missing. Traditional electronic video stabilization algorithms deal with the problem of missing video edge information through cropping. The clipping amplitude is determined by the intensity of video jitter and video smoothing algorithm. To solve this problem, this paper proposes a method to repair the video edge based on panoramic stitching. Use the reciprocity of jitter to segment the video; combined with the video motion trajectory, the extraction range of feature points is reduced; perform homography transform on each frame in the segmented interval, and then stitch it into an image containing panoramic information; finally, using the panoramic image, each frame after offset is stitched to repair the video with missing edge information. The experimental results show that using the real scene information to restore the missing part of the video, the filling is natural, the effect is stable, and can be applied to a variety of complex environments.