Lens distortion is coupled with intrinsic and extrinsic camera parameters in traditional camera calibration algorithms. An algorithm which separates lens distortion from camera parameters is proposed. The algorithm is based on the fundamental property that a camera follows the perspective camera model if and only if the projection of every three dimensional (3D) line in space onto the camera plane is a line. For undistorted images, the cross product of collinear vectors which consist of any two feature points on a line should be a zero vector. Distortion parameters are found by using nonlinear optimization algorithms and the coupling between distortion center and distortion coefficient is also discussed. An evaluation method for the effect of distortion correction is designed so as to prove the correctness of the proposed algorithm. Once distortion correction is completed, intrinsic and extrinsic camera parameters can be solved linearly. The experimental results indicate that the proposed algorithm can implement the solving of all camera parameters by using only one calibration image. As a result, the efficiency of calibrating is improved. Moreover, the proposed algorithm performs more stably and has nearly the same accuracy as traditional methods.