Robust digital image stabilization using feature tracking

This paper presents a new robust digital image stabilization (DIS) algorithm which uses well-tracked feature points to estimate the motion between two consecutive frames. The motion prediction with the Kalman filter is incorporated into the Kanade-Lucas-Tomasi (KLT) tracker to further speed up the tracking process. A new scheme is proposed to adaptively update the Kalman filter. Simulation results show that the proposed algorithm can speed up the tracking process and obtain more stabilized performance.