An algorithm is presented for determining the three-dimensional 3-D motion of a rigid object from three time-sequential perspective views taken by a stationary camera. In solving the problem, six or more line correspondences over three frames are used. It is found that the rotation and translation can be solved separately. Iterative methods are needed for solving the rotation parameters but linear algorithms are available for the translation parameters. For the special cases of constant rotation, only five line correspondences over three frames are needed. Some simulation results are shown for this special case.