Spherical-projection-model-based self-calibration of rotation cameras

A novel spherical-projection-based self-calibration scheme for a pure rotation camera is proposed in the paper. Specifically, firstly, the property is analyzed that the distance between spherical projection points that relate to space static ones are unvaried when the camera in the state of pure rotation motion; then, using this property, the constraint equations set that with respect to camera intrinsic parameters is constructed, and solved by the nonlinear least squares algorithm. Compared with the existing self-calibration methods, the proposed scheme can obtain 4 intrinsic parameters by straightly utilizing the correspondence points that on only two images. Then this method can avoid complex matrix computation, and 4 matching points is sufficient for accomplishment the 4-parameter calibration task as while as online and offline requirements are both satisfied. Experiments results verify that this method has a good robustness both for the image noise and the presence of the translation.