A novel energy model based predictive motion estimation algorithm
Amir Ghahremani, Amir Mousavinia · 2014
Motion estimation plays a vital role in video compression and many algorithms have been introduced to implement it. PSNR and computational complexity are two major factors comparing these algorithms. In general, algorithms with high PSNRs are most often suffer from excessive computational overhead. Among the existing ideas, predictive motion estimation recently has been considered as a prone to improve the challenging trade-off between PSNR and computational complexity. In this regard, this paper proposes a novel Energy Model based Predictive Motion Estimation (EMPME) algorithm, which benefits from the energy histogram of image blocks in addition to the previously found motion vectors to earn a more precise and fast prediction. In comparison with others, our approach eventuates into more accurate prediction by emphasizing on blocks with higher dynamic similarities. Simulation results show that the proposed technique is not only 35% faster than TSS (Three Step Search) algorithm but also it has improved the PSNR value.