A novel predictive global motion estimation for video coding

Global motion compensation improves the coding efficiency by reducing the transmitted motion vectors and reconstructs a more accurate prediction image for texture coding. Predictors are widely in local motion estimation and it is well proven that the use of predictors can speed up the local motion estimation process. In this paper, we propose a novel predictive global motion estimation algorithm, which combines the prediction idea and global motion estimation. The proposed algorithm first examines sum absolute difference of a set of candidate predictors and uses the predictor: with minimum SAD as the initial search center of N-dimensional diamond search to find the final global motion parameter. Although the PSNR of proposed algorithm is slightly lower than MPEG-4 VM GME, the speed of the proposed algorithm is significantly faster.