一种综合搜索策略的快速运动估计算法*

提出了一种综合搜索策略的运动估计算法。该算法首先采用中值预测提前终止判断策略，然后基于块运动类型确定搜索起点，最后采用小十字模板与基于块的梯度下降搜索法(BBGDS)相结合的方法进行局部搜索。搜索过程中多处引入提前终止策略，进一步提高搜索速度。通过与综合性能代表当前国际先进水平的运动矢量场自适应搜索法(MVFAST)进行对比实验发现，该算法在基本保持搜索精度的情况下，有效提高了搜索速度，对于运动较大序列速度提高尤为明显，可以达到20%48%。

Fast motion estimation algorithm based on hybrid search strategy

This paper proposed a hybrid search strategy algorithm, which firstly adopted median prediction early termination strategy, then decided initial search point based on block motion type, finally searched locally by integrating unit-size rood pattern and block-based gradient descent search (BBGDS) algorithm. Incorporated early termination strategy in the method several times, further speeding up the search. Experimental results show that the proposed algorithm can improve search speed with similar search accuracy compared with motion vector field adaptive search technique (MVFAST), which represents advanced international standard. Especially for sequences with large motion content, the speedup could achieve 20%48%.