Optimization of UMHexagonS algorithm for AVS+
Shuai Wang, Yonglin Xue · 2015
Motion estimation (ME) plays an essential role in video coding, but ME is also the most time-consuming part. Therefore, the algorithm for ME should be optimized for guaranteeing the real time performance of encoder. UMHexagonS algorithm, as one of the fast ME algorithms, has been accepted by AVS+. An improved UMHexagonS algorithm is proposed in this paper. In the proposed algorithm, the fixed asymmetrical cross template is replaced with a dynamic asymmetrical cross template. Search points of multi-hexagon-grid search are reduced. 5×5 spiral search as well as the extended hexagon template is improved. Furthermore, a small square template is added in the last of the algorithm. The simulation results show that with the proposed algorithm, time consumed in ME can be reduced from 8.93% to 15.27% of the total encoding time while the coding quality of the UMHexagonS algorithm is guaranteed.