Global motion compensation for image sequences and motion object detection

A effective approach to global motion compensation and moving object extraction is proposed. Firstly, feature-block detection and search-matching algorithm are used to attain background motion vectors of global image. Then, once the global motion vectors are robustly estimated, relatively stationary background can be almost completely eliminated through the registration difference algorithm. Finally, higher-order statistics is used to attain the motion target exactly. The experimental results validate that the proposed algorithm improves the performance of detection moving targets and effectively restrains the noise disturbance.