An adaptive cross search algorithm for block matching motion estimation

Motion estimation plays an important role in the motion compensated video coding framework. Due to the high computational complexity of the exhaustive search, many sub-optimal fast search algorithms, aiming to achieve the best trade off between distortion and search speed, are proposed. We observe that the distortion gradient of a search point on the block distortion surface (BDS) monotonously decreases with increasing distance from that point to the global minimum point. Based on this property, we propose a novel adaptive cross search (ACS) algorithm that can distribute the computation powers over the search space efficiently. Simulation shows that ACS achieves competitive reconstruction visual quality as well as reduced computational complexity.