Abstract: In this paper, instead of corresponding image points which are widely used in the literature, corresponding image conics are used to calibrate a rotating camera. An image conic can be the projection of either a planar conic or silhouette of a conicoid. We show that at least three images, taken from the same position with different camera orientations, of two planar conics, or two conicoids, or one conic and one conicoid, are needed for the calibration, and a closed-form solution can be obtained. Our new conic based calibration technique requires no knowledge of camera's orientations, and relies solely on image correspondences. Since a conic contains more useful information than a line or a point, it is much easier to establish feature correspondences. In addition, it is also helpful to improve the robustness of calibration. Both computer simulation and experiments with real images are done to validate our new technique.