搜索模式自适应快速运动估计算法

Fast Motion Estimation Algorithm of Search Pattern Adaptation

提出了一种搜索模式自适应快速运动估计算法(PAFME),首先利用变块尺寸运动估计的特点与运动矢量的时空域相关性,预测初始搜索中心;采用多种搜索模式以适应不同运动特征,提出了搜索模式自适应的选择机制,以节省不必要的搜索点加快搜索速度;又能避免陷入局部极小。实验结果表明,与H.264/AVC的参考软件JM12.4相比,该算法使整像素精度运动估计的速度提高了30%～40%,同时保持了图像质量和码率基本不变。

A search Pattern Adaptive Fast Motion Estimation algorithm （PAFME） is proposed. According to the characteristic of variable block size motion estimation and temporal -spatial correlation of motion vectors, initial search centers is predicted firstly. Secondly, diverse search patterns suit to different motion features and adaptive pattern selection mechanisms are employed in order to skip unnecessary search points, improve search speed and avoid dropping into the local-minimum. The experimental results show that, compared with the H.264/AVC reference software JM12.4, the proposed algorithm can achieve up to 30%-40% speed up ratio with a little PSNR loss and not much noticeable bit-rate increment.