| Video image stitching is an important application in the field of computer vision research,and has been widely used in military,medical,daily security surveillance and other fields.With the rapid development of information science and technology,various camera devices have been widely used in people's production and living,such as video surveillance cameras everywhere.Any camera has a range of viewing angles.Under a certain focal length,the field of view of the captured images is limited,and it is difficult to meet the requirements of people who want to obtain a large-view panoramic image in certain scenes.In order to obtain a larger field of view,stitching images from different angles taken by the same camera or stitching images taken by different photographic equipment is currently widely used.Obtaining large-view panoramic images through image stitching has very important research significance and value.In the video image stitching process,two crucial steps are video image registration and video image fusion.By studying the existing literature,this paper proposes its own improvement methods for these two key steps.The key research work and innovations of this paper can be summarized as the following points:(1)An improved ORB image matching algorithm based on particle swarm optimization is proposed.First,the improved algorithm selects the ORB algorithm with fast rate and high matching accuracy for feature point detection.In the search phase,particle swarm optimization is used to optimize and eliminate false matches.After experimental verification,compared with the original registration algorithm and the algorithm improved by the RANSAC method,the improved algorithm can effectively reduce the amount of mismatches and improve the accuracy of finding registration point pairs.The improved algorithm not only improves the correct matching rate,but also has high adaptability in various scenarios.(2)An improved image stitching method based on seed region growth algorithm and Poisson fusion is proposed.The robustness of the image fusion algorithm has an important impact on the final spliced image quality.The gray difference algorithm is added to the seed area growth algorithm to optimize and find the best stitching;then the image is stitched together with the Poisson fusion algorithm.Experimental results show that the improved algorithm can largely eliminate ghosting and stitching in the final image stitching result,and reconstruct a panoramic image close to the real scene.Theimproved algorithm can achieve better robust stitching results in different scenarios.(3)The improved image registration and image fusion stitching algorithm is applied to the video image stitching process.Compared with the existing video image stitching algorithms,the experimental results show that the improved algorithm achieves better stitching effect in the stitching of multi-camera video images.The research of this paper is to improve the effect of video image stitching,and the real-time performance of the algorithm is the next research focus. |
