Research on Video Stabilization Algorithm Based on SIFT and Improved RANSAC

Abstract

The video gets from the device should be stabilized by removing the jitter caused by non-smooth movement of the platform. To solve the problem, Firstly using the SIFT algorithm to extract the feature points of the video images, and then using the improved Random Sample Consensus (RANSAC) algorithm to obtain the motion parameters. To improve the speed of computation, a method of pre-detection is used to discard those temporary models that were not candidate models. To eliminate the outliers, a random block selecting method was used to select samples, which improves the precisions. The experiment result shows that this algorithm largely reduces the amount of computation and improves the speed of video stabilization when the accuracy do not change notably. thereby effectly enhances the video stabilization.