In this paper, a novel and efficient block-matching criterion called mean absolute error of projective sum (MAPS) is proposed to significantly reduce the computational complexity of block-based motion estimation. It can save about 78% computational load in comparison with that of the conventional mean absolute difference (MAD). Simulation results show that its prediction quality is close to the MAD's. By applying the partitioning technique in MAPS, the prediction quality can be further improved such that the proposed criterion can be applied in different applications for motion estimation. Further, according to its simple and regular properties, it is very suitable for VLSI implementation.
展开▼
