A center-biased adaptive search algorithm for block motion estimation

Abstract

A number of sub-optimal, but faster, search algorithms have been proposed in the literature, in order to alleviate the complexity associated with motion estimation by the full-search method. A new sub-optimal center-biased adaptive search algorithm for motion estimation is proposed; we refer to this algorithm as center-biased dynamic MInima Bounded Area Search (MIBAS) algorithm. The novelty of MIBAS is the checking point pattern at each subsequent step, composed of points lying in the area bounded by two local minima found at the present step, rather than points lying in a small neighborhood around a local minimum. The simulation results show that, compared to other fast algorithms, the center-biased MIBAS is more robust and produces smaller prediction errors and more reliable motion vectors, while it has a comparable computational complexity.