Elastic Warping with Global Linear Constraints for Parallax Image Stitching
Feiyang Xue, Danchen Zheng · 2023
Image stitching is used to produce high quality panoramas, meaning that overlapping areas are perfectly aligned and non-overlapping areas do not produce noticeable distortion. Achieving this goal is difficult when parallax is present. Elastic warping can be a picture exactly aligned at the location of feature points, but it can produce obvious artifacts since it only takes local alignment into account. In this paper, we propose a post-processing method of elastic warping to solve this problem. We use line segments as global constraint for the warping and deform the result of the elastic warping once more. The experiments show that our method can greatly alleviate the artifacts of elastic warping in scenes with large parallax and also has well-performance compared with other methods.