Image Stitching with Weighted Elastic Registration
Kang Sun, Yi Wan · 2018
In order to obtain high-quality image stitching results with parallax tolerance, some superb image stitching techniques have been proposed in recent years. One of the methods is to use the approximating thin-plate splines to register the image. However, in some cases with large parallax, it is easy to produce visible distortion at the edge of the overlapping region, making the result look unnatural. Based on the careful study of image stitching with TPS model, this paper proposes a weighted deformation scheme, which effectively reduces the distortion of the stitched image nearby the overlapping region and further improves the visual effect of image stitching. Experimental results show that the proposed method has achieved satisfactory results in several challenging cases.