Fast Multi-level Successive Elimination Algorithm for Motion Estimation in H.26L

Abstract

Motion estimation plays an extremely important role in the video coding. The objective of the motion estimation is to remove the temporal redundancy between video frames so that the motion compensated frames can be coded efficiently.
H.26L video coding is the most efficient coding standard currently available. It uses multi-mode with variable block-size motion estimation to improve the accuracy. However, the conventional full search algorithm will be a heavy computational load in this situation. To reduce the complexity, we propose a fast multi-level successive elimination algorithm (FMSEA) for H.26L multi-mode motion estimation search. The proposed method is mainly based on the combination of a modified multi-level successive elimination algorithm (MSEA) with a motion refinement approach and a half-stop decision that skips the 8x4, 4x8, and 4x4 sub-block motion searches. Experimental results show that FMSEA is very efficient in terms of the computational speedup and video reconstruction quality for H.26L.