An improved motion vector estimation algorithm was proposed to remove effectively the inter frame jitter of the video sequences recorded by airborne imaging equipment and to realize the electronic image stability. The BRISK operator and the corresponding feature extraction model were introduced. Then, the BRISK operator was employed to extract and match the stable feature points of neighbor frames. By combining with the affine transformation motion model, the rough estimation of inter frame motion vectors was implemented. To improve the accuracy of motion vector estimation, the particle filter with the weighted least-square method was employed to solve the problem of the motion vector calculation inaccuracy caused by different depths of the feature points. Finally, the Kalman filter was used to separate the jitter components from the global motion vectors for compensating the video sequences frame by frame. The experimental result indicates that the inter-frame transformation fidelity (ITF) of video sequence has improved about 3 db, which verifies that the proposed stabilized algorithm removes inter-frame jitter of high frequency under complicated movement with fast speed, higher accuracy and stronger robustness.