基于MAP的全局运动估计算法  

Global Motion Estimation Algorithm Based on MAP

作　　者：刘征宇[1] 周孝宽[1] 

机构地区：[1]北京航空航天大学宇航学院,北京100083

出　　处：《计算机工程》2009年第3期237-239,共3页Computer Engineering

摘　　要：常见的全局运动算法存在的问题是仅仅考虑像素的残差,而忽略了相邻像素和分割信息等因素对算法的影响。为了解决上述问题,该文提出一个基于最大后验概率(MAP)的全局运动估计算法。通过MAP算法定义系统能量,引入先验能量来描述像素的相邻关系和分割信息,将系统能量作为目标函数通过LM算法估计全局运动参数。实验结果证明了该算法的有效性。One of the main problems related to global motion estimation is that the algorithm only considers the change information in the difference image, but the import information of segmentation and correlation has been ignored. To overcome such drawbacks, a technique of global motion estimation based on MAP is proposed. The system energy is defined through MAP and the potential energy is defined by segmentation mask and the prior information. Using the system energy as object function, the global motion is estimated. Experimental results confirm the effectiveness of the proposed technique.

关 键 词：全局运动估计 最大后验概率 Gibbs能量 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]