详细论述了基于块匹配的鲁棒运动估计算法。跟已有的基于块匹配的运动估计算法比较，首先，我们引入颜色信息来提高运动估计的准确性；其次，在更广泛的意义上运用自适应策略来减少计算量并同时保证算法的鲁棒性；最后，提出的基于预测修正的复合查找方法充分利用了物体运动的全局信息，克服了三步查找算法以及全查找算法的缺点并充分发挥它们二者的优点从而提高查找的效率和匹配精度。实验结果表明基于块匹配的鲁棒运动估计算法具有抗干扰能力强、运动估计准确、计算效率高等优点。

We proposed a robust approach to object motion estimation between frames based on block matching in this paper.Compared with existing methods for object motion estimation between frames,our approach,firstly,employs color information to improve accuracy of block matching;secondly,by adopting a wide range of adaptive scheme,the robustness and high efficiency of the proposed approach can be achieved; finally,prediction correction based compound search algorithm integrates history information of object motion together with advantages of three step search algorithm andfull search algorithm to yield efficient andaccurate motion estimation.Experiment results demonsrate that the proposed approach can track objects moving in a very complex background or that in a low SNR environment,so our approach can achieve more robust motion estimation thanexisting block matching based methods.