Real-Time Optical flow-based Video Stabilization for Unmanned Aerial Vehicles

• DOI: 10.48550/arxiv.1701.03572
• Publication Date: 2017-01-01
• Authors: Anli Lim; Bharath Ramesh; Yue Yang; Cheng Xiang; Zhi Gao; Feng Lin
• Keywords: Optical Flow; Benchmark (surveying); Image stabilization; Tracking (education)

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