Video stabilization algorithm based on background feature point matching

Abstract

Global motion estimation between two successive frames is important to the process of video stabilization. In the proposed approach, the estimation of global motion was based on the background feature points (BFPS). First, feature points were collected from the input video by FAST operator; second, feature point's descriptor and matching were based on FREAK operator.The MSAC is used to classify the BFPS. Last, the six parameters of the affine transform model to calculate the interframe motion estimation vector. The experiment results show that he proposed method can stabilize inter-frame jitter, in the meanwhile, it improve the video quality effectively.