基于H.264快速运动估计算法研究

On Fast Motion Estimation Algorithm Research for H.264

为了降低运动估计的计算复杂度,提出了一种基于H.264的快速运动估计算法,该算法使用了候选尺寸块的预测和两种匹配模板的搜索,通过部分像素匹配、多参考帧选取、亚像素匹配来获取最佳运动矢量。实验表明,与FS算法和UMHexagonS算法比较,该文所提算法在保持了相近的峰值信噪比和码率的情况下,运动估计耗时平均下降了69%和13%,编码速度大幅提高。

A novel fast ME algorithm for H. 264 is proposed in this paper in order to reduce the computation of motion estimational （ME） module. The proposed algorithm adopted candidate size block prediction and two kinds of matching template searching, the best vector is achieved by matching part pixel, selecting reference frames and matching sub-pixel. Experimental results show that it has saved 69 % and 13 0% on average of motion estimation time compared to FS and UMHexagonS respectively. The proposed algorithm can highly speed up video coding with similar PSNR（Peak Signal Noise Ratio）and code rate.