A stratified self-calibration method for circular motion in spite of varying intrinsic parameters

Self-calibration for imaging sensors is essential to many computer vision applications. In this paper, a new stratified self-calibration method is proposed for a camera undergoing circular motion with varying intrinsic parameters. It is shown that the plane at infinity in a projective frame can be identified by (i) known rotation angles between any two views and (ii) square pixel assumption of the camera. Once the plane at infinity is identified, the calibration matrices of the cameras and the upgrade to a metric reconstruction can be readily obtained. The proposed method is rather flexible in that it allows not only the focal length but also the principal point to vary from image to image. Experimental results for both synthetic data and real images are provided to show the performance of the proposed method.