Motion compensation algorithms are compared and analyzed, including Kalman filter algorithm, average filter algorithm, Gaussian kernel filter algorithm, parabolic fitting algorithm, and damping coefficient algorithm. Motion estimation vectors of several fluctuated video are obtained by Laplacian bit-plane method, which are utilized to test the above compensation algorithms. The conclusion for the preceding five algorithms conduces to select appropriate motion compensation algorithms in video stabilization. It is verified that damping coefficient algorithm is a quite effective algorithm, except that its damping coefficient can not be adjusted automatically. An improved adaptive damping coefficient algorithm is presented, which is more robust and effective.