Video Stabilization Technique for Thermal Infrared Aerial Surveillance

Video taken from Unmanned Aerial Vehicles (UAV) frequently suffers by unwanted motion of the sensors, which severely affects the performance of automatic target detection and tracking systems. In this paper we present a new video stabilization algorithm for thermal infrared videos captured by UAV to remove unwanted motions and produce stabilized video frames. Initially, we use Scale Invariant Feature transform (SIFT) for key point detection and matching between successive frames. Then, affine transformation model is used to estimate the global motion parameters between two successive frames. After that, the undesired motions are compensated and spatiotemporal filtering is used to remove the noises in the video. Finally, all frames are transformed to obtain stabilized video frames. Experimental result shows the efficiency of proposed algorithm in terms of quantitative and qualitative aspects.