Naturalness-Preserving Image Stitching Based on Optimal Seam Estimation Considering Parallax
Yuya Nagasugi, Ryosuke Harakawa, Masahiro Iwahashi · 2020 IEEE 2nd Global Conference on Life Sciences and Technologies (LifeTech) · 2020
Image stitching that produces a panoramic image from multiple images is one of the fundamental techniques in computer vision. Conventional image stitching methods have a problem that they yield a visually implausible result when parallax, i.e. the visual difference between the same objects in multiple images because of the differences of camera locations and angles, exists. To solve this problem, this paper proposes an image stitching method based on optimal seam estimation considering parallax. Compared with existing methods, the novelty of this work can be found in the use of optical flow and color information for estimating the optimal seam. Concretely, parallax is represented as the movement of the same object between the input two images. Then, the movement is estimated by using optical flow. Furthermore, by newly using color information as well, the proposed method can accurately estimate the movement even for the region with small gradients. Experimental results show that the proposed method can produce a naturalness-preserving stitching result.