Abstract: The macroblock motion estimation process is a long time consuming part in H.264 standard. Aiming at this problem, a fast mode decision algorithm for motion vector prediction is presented in this paper for decreasing it efficiently. The average of predictive motion vector is computed according to the spatial correlation of video sequence, and the current block as a slow or fast motion block and the direction of motion vector are determined and an appropriate search pattern is selected according to the square mixed pattern. It is decided that whether the motion estimation terminated or not according to a given threshold. Experimental results show that the proposed algorithm can decrease 36.34% of average motion estimation time than the existing EPZS algorithm does with about the same quality of video coding.

Key words: H.264 standard, motion estimation, square mixed model, adaptive early termination

摘要： 针对H.264标准中宏块运动估计时间长的问题，提出一种运动矢量预测的快速模式选择算法。根据视频序列的空间相关性，求出平均预测运动矢量，通过正方混合形模型确定运动量大小和运动矢量的方向特性，选择合适的模型进行搜索。根据设定的阈值，判定运动估计是否提前中止。实验结果表明，与EPZS算法相比，该算法在保证视频编码质量的同时，平均能减少约36.34%运动估计时间。

关键词: H.264标准, 运动估计, 正方混合形模型, 自适应提前中止

CLC Number: 

• TN919.81