Abstract:

Camera poses play a crucial role in stitching overlapped images captured by the camera to achieve a broad view of interest. In this paper, we proposed a robust camera pose estimation approach to stitching images of a large 3D surface of known geometry. In particular, given a collection of images, we first construct matrices of relative camera poses, where each entry is achieved by solving a perspective-n-point (PnP) problem over its corresponding pair of images. We then jointly estimate camera poses by solving an optimization problem that exploits the underlying rank-2 matrix of relative poses and the joint sparsity of camera pose errors. Lastly images are projected to the 3D surface of interest based on estimated camera poses for further stitching process. Numerical experiments demonstrate that our proposed method outperforms existing methods in terms of reducing camera pose errors and improving PSNRs of stitched images.