Rapid block-based global motion estimation and its applications

This paper proposes a novel low-complexity block-based global motion estimation algorithm for real-time digital video processing applications. The algorithm involves an 8/spl times/8 block-based local motion estimation followed by a least square method to obtain the global motion vector parameters. The translation-zoom global motion model with 3 parameters was adopted and least square method computation is done after removing outliers. Blocks are considered outliers if: (i) the block is over-cluttered or (ii) the reference block has too little activity for an accurate motion vector to be estimated. The algorithm is used to estimate the global motion of 3 QCIF test streams at 10 fps and the parameters are used in 5 typical applications Respective simulations have produced favorable results emphasizing the usefulness of the proposed algorithm.