Research On Video Stiching To Solve Ghosting And Misalignment

Video stitching is an important means to achieve panoramic video.In the era of increasing shooting capabilities of mobile devices,stitching videos captured by hand-held mobile cameras can essentially enhance entertainment experience of ordinary users.However,such videos usually contain heavy shakiness and large parallax.These factors will affect the stitching effect.A novel joint video stabilization and video stitching algorithm is proposed for stitching jitter videos,which can effectively remove jitter and ghosting and misalignment,and improve the video perception after stitching.The video stitching algorithm proposed mainly includes three stages:1.Because video stitching and video stabilization are very dependent on feature matches,in order to improve the matching accuracy of feature matches.An improved image registration method is proposed for video stitching: firstly,videos will be preprocessed and overlapping regions of video frames will be extracted adaptively;then SIFT algorithm is used for feature matches;and finally an improved RANSAC algorithm is proposed.The main steps of the improved RANSAC algorithm are: first the distance-based filtering algorithm is used to filter invalid feature points,then a fast model-checking algorithm is used to quickly stop the abnormal model iteration,and finally a grid-based partition sampling algorithm is used to select better sample points in iterations.2.In order to solve the problem of ghosting and misalignment of people in overlapping regions,an improved finding seamline algorithm that can bypass the human body is proposed:first the human body region can be extracted through HOG;then the proposed energy function to re-establish the energy map of the video frame is established;and finally the minimum flow algorithm is used for finding the improved seamline.3.The main component of the method is the optimization algorithm of joint video stabilization and video stitching.The algorithm is to optimize video stabilization and video stitching at the same time instead of calculating video stabilization and video stitching independently.Thus stable mosaic videos can be obtained by the algorithm proposed.Through the simulation of walking and mobile shooting scenes with a hand-held motion camera,the experiment proves that the proposed algorithm has good performance in test data in the ghosting problem of the overlapping area and has achieve good result of video stability and the actual stitching result.