根据序列图像中运动矢量的中心偏置分布特性和矢量间的时空相关性,设计了一种基于连续消除的正方形--菱形搜索策略。在此基础上,融合了搜索起点预测和自适应搜索等技术,提出了一种可预测搜索起点的自适应的基于连续消除的正方形--菱形搜索算法(PA-SEA-SDS)。实验表明,该算法以较小的代价取得了与全搜索算法相当的效果,在搜索次数和精度上均优于菱形和正方形-菱形搜索算法。 According to the central bias distribution of motion vectors and the spatiotemporal correlation between vectors, a square-diamond search strategy based on continuous elimination is designed. Based on this, the technology of search starting point prediction and adaptive search is integrated, and an adaptive continuous square-diamond search algorithm (PA-SEA-SDS) which can predict the search starting point is proposed. Experiments show that this algorithm achieves the same effect as the full search algorithm at a small cost, and is superior to the diamond and square-diamond search algorithms in terms of the number of searches and the precision.