Fast motion vector estimation using multiresolution-spatio-temporal correlations

We propose a new fast algorithm for block motion vector (MV) estimation based on the correlations of the MVs existing in spatially and temporally adjacent as well as hierarchically related blocks. We first establish a basic framework by introducing new algorithms based on spatial correlation and then spatio-temporal correlations before integrating them with a multiresolution scheme for the ultimate algorithm. The main idea is to effectively exploit the information obtained from the corresponding block at a coarser resolution level and spatio-temporal neighboring blocks at the same level in order to select a good set of initial MV candidates and then perform further local search to refine the MV result. We show with experimental results that, in comparison with the full search algorithm, the proposed algorithm achieves a speed-up factor ranging from 150 to 310 with only 2-7% mean square error (MSE) increase and a similar rate-distortion performance when applied to typical test video sequences.