Abstract: Camera self-calibration is one of the fundamental issues in computer vision. This paper discusses the calibration of camera intrinsic parameters from the rotation of camera. When camera is rotated around its coordinate axis, the formulas of camera intrinsic parameters are obtained by the algebraic method. These formulas are very useful when 2D projective transformation is very close to the theoretic P. When rotational axis are unknown, the general solution of camera intrinsic parameters is obtained by the eigenvectors of 2D projective transformation. The camera intrinsic parameters could be uniquely determined through the method based on two different rotational axes. These results are the theoretic basis of the self-calibration algorithm. Meanwhile a practical algorithm is provided. Tests with synthetic data and real images indicate that the algorithm presented in the paper is robust.