A novel fast sampling predictive block match algorithm

Block-based motion estimation is essential in video coding applications, which can diminish the temporal redundancy of sequences effectively. Though full search algorithm (FSA) can undoubtedly get global optimum estimation, it is computationally intensive. In this paper a novel algorithm named as the Sampling Predictive Block Match Algorithm (SPBMA) is presented, which combine the ideas of motion vector mean predictor, sampling match and adaptive diamond search. Our experiments verify the superiority of the proposed algorithm in precision and rapidity (for the Mother & daughter sequence, SPBMA is 108 times as fast as FSA with only 0.75% lower in PSNR). It is suitable for the wireless video communication with the rigorous demand for ultra low power.