一种基于区域搜索的全局运动估计与补偿算法

开展全局运动估计与补偿研究是进行动态目标检测中的基础和前提。在总结现有运动估计与补偿方法的基础上,提出一种基于图像分割区域的运动性和大小的全局运动估计与补偿算法。首先,通过建立区域搜索的全局运动模型,同时进行区域定性分割和区域大小排序;然后,根据误差最小化准则在指定的分割区域中进行线性递归搜索,利用门限准则寻找出最佳的运动估计参数;最后,根据双线性内插法获得运动位移量。实验结果比较可知,所提算法较三步搜索算法(TSS)和全局搜索算法(FSA)等传统算法具有更高的准确性(图像平移帧差)和实时性(算法运行时间),能够很好地实现运动背景的全局校正。

A Novel Global Motion Estimation and Compensation Algorithm Based on Region Search

Researching of the global motion estimation and compensation is the basis of dynamic target detection.On the basis of analyzing the traditional methods,a novel algorithm is proposed based on the motility and the size of image segmentation regions for the global motion estimation and compensation.Firstly,the global motion model of region search is built,and the original image is segmented qualitatively and these regions are sequenced from the bigger to the smaller.Secondly,the min error(ME) rule and threshold rule are used to find the optimal motion estimation parameters in the segmentation region based on linear recursion search method.Thirdly,the bi-linear inner-interpolation is used to acquire the motion displacement vector.Comparative experiments show the proposed algorithm has better veracity(displace frame difference) and real time(running time of algorithm) than the traditional methods(TSS and FSA),and the algorithm has better quality in motive background global revision.