Abstract: A fast global motion estimation algorithm is proposed by employing the gyro with ordinary precision to sense the camera status. Firstly gyro is used to sense the jitter of camera system, and the bearings (quadrant) of video jitter are achieved from the gyro output. Secondly, singular points of local motion vector are filtered using the acquired direction and position as the reference. Finally, global motion parameters are worked out using the least square method (LS). Experimental result shows that this algorithm has the advantages of fast operation, high accuracy of motion estimation and low precision requirement of gyro. Also it could effectively filter singular points stemming from either interference of local motion or error of calculation.