Self-calibration using two same circles

Camera calibration is a fundamental step in 3D reconstruction and computer vision. Considering the easy accessibility of two same circles, a method for self-calibration based on two same circles is proposed. The proposed method does not need any prior knowledge and known camera parameters. By taking three photos of the object containing two same circles from different views, the close solution for intrinsic and extrinsic parameters are first obtained by the invariance of circular points and common tangent of two circles. Then the solution is refined by the nonlinear optimization. This method could get the intrinsic parameters as well as the extrinsic parameters without complicated matching process. Extensive results show the accuracy, robustness and wide applications of the proposed method.