Robust and fast global motion estimation oriented to video object segmentation

Most global motion estimation (GME) methods are oriented to video coding while video object segmentation either assume no global motion (GM) or directly adopt a coding-oriented method to compensate GM. This paper proposes a hierarchical differential GME oriented to object segmentation. A combination of 3-step search and motion vector prediction is proposed for initial estimate. Two robust estimators are also proposed: to estimate global motion in the first frame and to reject outliers using object information. Subjective and objective results show that the proposed method is more robust and faster than the reference methods.