Research Video Stitching Method Based On Depth Information

With the rapid development of multimedia technology,multi-view video stitching technology is more and more widely used in large-scale scene monitoring system,security system,military system and biomedical system,which has broad research significance.After reading a large number of related literatures about video image stitching at home and abroad,it is found that when multi-view video stitching is carried out,the depth information between different frames changes because of the complex structure of the depth in the scene.Using the registration method in image stitching,the depth distortion problems such as artifacts and discontinuities will occur during stitching,which seriously affects the quality of large scene images.Moreover,due to the existence of depth distortion,when the camera moves or there are moving objects in the scene,the phenomenon of depth distortion will be strengthened,which will further blur and ghost the stitching large scene video image.Aiming at the problem of depth distortion in multi-view video stitching,this paper improves the original method of depth information acquisition,and reduces the adverse effects of depth distortion in video image stitching by adding depth information constraints to existing stitching methods.The main contents include the following parts:1.In video image stitching,the depth information between different frames in the scene captured by the camera is variable,so it is easy to appear depth distortion phenomena such as artifacts and discontinuous scenes.Moreover,when the camera moves or there are moving objects in the scene,the existence of depth distortion will further blur and ghost the mosaic large scene video image.Aiming at the problem of depth distortion,this paper studies the method of depth information acquisition,and proposes a spatial stereo matching algorithm based on the combination of fast prior parallax estimation and posterior parallax,which extracts depth information by fusing the temporal information refinement of video image sequence.2.Aiming at the problem of depth distortion caused by different depth sizes of the same scene point captured by two cameras in the two imaging planes,this paper proposes to add depth information to optimize and improve Automatic Stitching method,As Projective As Possible method,Shape Preserving Half Project method,and further restrict feature point selection by using depth information.One step to reduce the depth distortion in the stitching process.Experiments show that the proposed method can effectively improve the depth distortion caused by tilt and blur after video image stitching,and achieve high quality video image stitching.3.In the process of multi-view video image stitching,because the baseline of the camera is not on the same horizontal line,the baseline problem between different cameras needs to be considered before stitching.In this paper,the method of multi-view panoramic construction of wide-baseline image is improved by adding depth information,and the image of wide-baseline scene structure between image and non-planar image is carried out.The image stitching of multi-view input can be better realized to form a high-quality large scene image.