A fast three-step search algorithm by the utilization of multilevel vector partial sums

Due to the high computational requirement of the full-search algorithm for block motion estimation, fast block motion estimation algorithms are needed for real-time implementations of the video coding standards. Recently, a three-step search algorithm for block motion estimation has been proposed in the literature. In this paper, a fast three-step search algorithm is proposed to further reduce the computational complexity of the three-step search algorithm with no loss of accuracy. By using a multilevel vector partial sums and lower bounds in the proposed algorithm, a large number of possible candidate motion vectors are discarded while still retaining the optimal motion vector of the three-step search algorithm. It is shown that not all the levels of partial sums and lower bounds are needed. A method to select these vector partial sums and lower bounds are also presented. Simulations of the proposed algorithm are carried out for various benchmark video sequences and the results demonstrate that the new algorithm can reduce the computational complexity of the three-step search algorithm by 20 to 60 percent with no loss of accuracy.