A Robust Video Stabilization Method Based on Feature Point Extraction and Warping

Abstract

In recent technologies there are various applications which include a camera joined to a moving platform, for example, cars, drones and Unmanned Aerial Vehicles (UAV). The moving platform may suffer from vibrations which may cause unwanted motion in recordings that can cause degradation of performance in various applications like surveillance, tracking and detection of object. Stabilization of video in various applications is an emerging research area nowadays. To remove the unwanted motion from video, the stabilization is necessary to preserve the important content present in the video. In this paper the feature points from recorded videos are detected and then these feature points are extracted and matched. The obtained feature points are smoothed by K means clustering, a mesh grid on every video frame is set up and every grid is warped by matching and comparing the features points, from original video frame with the smoothed and stabilized feature points. The reduced distortions in the video are estimated from various parameters. The efficiency of algorithm is compared in which the robust video stabilization algorithm based on feature extraction and mesh grid warping obtains better improvement in Inter-frame Transform Fidelity (ITF) factor than the traditional video stabilization algorithm.