To solve the problem that there will b e massive blank area after 2D digital image stabilization compensates global motion vectors,we propose a video stabilizing algorithm which can sm ooth feature trajectories quickly without using homography model to estimate global motion vectors.First,we extract images′ loc al feature points with improved SURF algorithm.Then,by connecting the matching points between frames based on consist ency of spatial motion,we get the feature trajectories.At last,we establish an objective function which takes both smoothness of the feature trajectories and degeneration of video quality into consideration to smooth the feature trajector ies and get the stabilized video.The results of experiment indicate that the stabilized video with this method has 30% less bla nk area and can satisfy human sense compared with those with Matsushita method,a nd can reduce the motion inpainting.Moreover,it eliminates the accumulated error between frames in motion estimation and it is helpful to the video with large local motion in foreground.