This paper describes the development of a novel algorithm to tackle theproblem of real-time video stabilization for unmanned aerial vehicles (UAVs).There are two main components in the algorithm: (1) By designing a suitablemodel for the global motion of UAV, the proposed algorithm avoids the necessityof estimating the most general motion model, projective transformation, andconsiders simpler motion models, such as rigid transformation and similaritytransformation. (2) To achieve a high processing speed, optical-flow basedtracking is employed in lieu of conventional tracking and matching methods usedby state-of-the-art algorithms. These two new ideas resulted in a real-timestabilization algorithm, developed over two phases. Stage I considersprocessing the whole sequence of frames in the video while achieving an averageprocessing speed of 50fps on several publicly available benchmark videos. Next,Stage II undertakes the task of real-time video stabilization using amulti-threading implementation of the algorithm designed in Stage I.
展开▼