Low Complexity Mode Decision for H.264 Based on Macroblock Motion Classification

The H.264/AVC achieves higher compression efficiency than previous video coding standards. However, the full RD cost calculations for all intra-prediction modes and exhaustive searches for optimal motion vectors for variable block sizes result in extremely high computation complexity, which obstruct it from practical use. In this paper, an efficient algorithm is proposed to reduce the complexity of macroblock mode decision. Firstly, the proposed algorithm is to identify the boundary region and the interior region of the motion object by using the motion vectors information. Secondly, the boundary region was classified into two types of regions by the coded modes information. After that we process the different region distinctly. Experimental results show that the algorithm can save the encoding time up to 68% on average compared to the conventional method in the JVT JM8.6 reference encoder at the cost of negligible performance degradation.