VLSI oriented fast motion estimation algorithm based on macroblock and motion feature analysis

Abstract

The latest H.264/AVC standard can provide us superior coding performance. However, the new technique also brings about complexity problem, especially in motion estimation (ME) part. In hardware, the pipeline stage division of H.264 based ME engine degrades many software oriented complexity reduction schemes. In our paper, we propose one VLSI friendly fast ME algorithm. Firstly, pixel difference based adaptive subsampling achieves complexity reduction for homogeneous macroblock (MB). Secondly, a multiple reference frame elimination scheme is introduced to early terminate ME process for static MB. Thirdly, based on the motion feature analysis, the search range is adjusted to remove redundant search points. Experimental results show that, compared with hardware friendly full search algorithm, our proposed algorithm can reduce 71.09% to 95.26%ME time with negligible video quality degradation. Moreover, our fast algorithm can be combined with existing fast motion estimation algorithms such as UMHexagon search for further reduction in complexity and it is friendly to hardware implementation.