Obtaining a motion vector for a partition of a macroblock in block-based motion estimation

摘要

A method of obtaining a motion vector for a partition of a macroblock in block-based motion estimation by dividing macroblocks into partitions and determining a partition motion vector for each partition. A best vector is selected for each partition from a set of candidate motion vectors comprising the vector from the macroblock and the vectors from neighbouring macroblocks. A score is calculated for the partition for each candidate motion vector and the vector having the best score is selected as the as the motion vector for the partition. The partitions may be partitioned into sub-partitions to select the best vectors for the sub-partitions. The process may be repeated until the sub-partitions are pixels to obtain the best vectors for each pixel. The score may be a sum of absolute differences.