Selective Multi-Resolution Motion Estimation Using Half-pixel Accuracy and Characteristics of Motion Vectors

摘要

A new multi-resolution motion estimation (MRME) algorithm using half-pel accuracy motion estimation (HPAME) and characteristics of motion vectors in the baseband is proposed in this paper. Generally, MRME method in the WT (wavelet transform) domain needs the exact motion vectors in the baseband because those are used as initial motion vectors in higher frequency subbands. Therefore the proposed method uses the HPAME to estimate the displacement exactly in the baseband. Then the displacement in the higher frequency subbands is selectively estimated based on the characteristics of the motion vector in the baseband. That is, the displacements in the higher frequency subbands are estimated only for the blocks which have motion vectors in the half-pixel accuracy for the corresponding block in the baseband. This is based on the fact that the block having the half-pixel accuracy motion vector in the baseband has a high probability to have a small displacement for the corresponding block in the higher frequency subbands. In the proposed method, by using the HPAME in the baseband and selective motion estimation in the higher frequency subbands, we can obtain motion-compensated image of the similar quality with the conventional method although we reduce the computational amount and the bit rate considerably.