Research and approach on video stabilization algorithm

This paper presents a video stabilization algorithm based on Scale Invariant Feature Transform and the improved RANdom SAmple Consensus (RANSAC) algorithm. Feature points of the video images are extracted using the SIFT algorithm, and then a compensation matrix is obtained using the improved RANSAC algorithm. Pre-detection is used to accelerate the computation by discarding those temporary models that are not candidate models. The experiment result shows that this algorithm largely reduces the amount of computation, as well as improves the speed of video stabilization without changing the accuracy notably.