A neighborhood elimination approach for block matching in motion estimation

A new algorithm has been proposed for reducing the number of search locations in block matching based motion estimation. This algorithm uses spatial correlation to eliminate neighboring blocks having low probability of being best match to the candidate block. Existing fast BMAs use a fixed pattern to find the motion vector of a macroblock. On the contrary, the proposed algorithm is independent of any such initially fixed search patterns. The decision to eliminate the neighborhood is taken dynamically based on a preset threshold Th  . The extent to which the neighborhood can be eliminated is configured using the shift parameter δδ. Thus, reduction in the number of search positions changes dynamically depending on input content. Experiments have been carried out for comparing the performance of the proposed algorithm with other existing BMAs. In addition, an Adaptive Neighborhood Elimination Algorithm (ANEA) has been proposed whereby the Th   and δδ parameters are updated adaptively.

► Statistically, in most sequences, motion information follow a local pattern.
► Our NEA exploits this locality preserving property of motion vectors.
► Based on the motion content, it adaptively eliminates neighborhood search positions.
► Empirical results on MEPackage and H.264 reference software.