Motion adaptive search for fast motion estimation

The paper presents motion adaptive search (MAS), that effectively utilizes motion activity information to achieve superior performance both in terms of quality and speed in comparison with other fast search algorithms for motion estimation. The main features of MAS are: the early elimination of search by detecting the stationary macroblocks; the adaptation of both the search window size and the search strategy based on motion activity information; the minimization of overlap of search points by performing intelligent search.