基于预测的方向性菱形快速运动估计算法

A Predictive Directional Diamond Fast Motion Estimation Algorithm

基于对运动矢量分布的方向性中心偏移特性和起始点预测线性搜索技术的研究,提出了一种基于预测的方向性菱形快速运动估计算法,在保持搜索精度的同时有效降低了计算复杂度。实验结果显示,与运动矢量场自适应搜索算法(MVFAST)[1]相比,在图像质量几乎不下降的情况下,平均减少0.6个搜索点数,搜索速度提高明显,综合性能优于MVFAST算法。

Based on the research of directional centre-based motion vector distribution and predictive lineal search, this paper proposes a fast motion estimation algorithm-a predictive directional diamond fast motion estimation algorithm （PDDFA） . The new algorithm can effectively reduce computational complexity but keeps search accuracy at the same time. Experimental results show PDDFA saves 0.6 points on average under no obvious loss of video quality in contrast to MVFAST. The comprehensive performances of PDDFA excel MVFAST.