A fast two-step search algorithm for half-pixel motion estimation

With the extensive application of integer-pixel fast search algorithms, the half-pixel full search consumes a significant portion of the entire motion estimation time. To reduce its computational complexity, a Two-Step Search (2SS) algorithm for half-pixel motion estimation is proposed in this paper. The 2SS algorithm is based on the fact that the SAD distortion function increases monotonically as the search location moves away from the sub-pixel minimum within the ±1 pixel neighborhood of integer-pixel search result. Simulation results show that the proposed algorithm is simple and efficient. It requires less than one half of the computation compared with that for the full half-pixel search while maintaining almost the same image quality and bit rate.