Robust and fast global motion estimation for arbitrarily shaped video objects in MPEG-4

Abstract

Global motion estimation is used for exploiting temporal redundancies in arbitrarily shaped video objects, which is computationally the most demanding part within the MPEG-4 standard. In this paper, we propose a robust and fast method for the estimation of global motion of video objects used in the MPEG-4. Predictors are widely used in block-based motion estimation and it is well proven that the use of good predictors can speed up the motion estimation process. The proposed algorithm makes use of the shape information (alpha-plane) of video objects to form a new predictor. From the experimental results, we conclude that the new predictor provides additional performance gains with reducing computational complexity of global motion estimation.