图像对准估计参数的不确定性分析  

Uncertainty Analysis of the Estimated Parameters of Image Registration

作　　者：罗红根 Luo Honggen(Finisar Shanghai Incorporated,Shanghai 200030,China)

机构地区：[1]菲尼萨光电通讯(上海)有限公司,上海200030

出　　处：《信息与电脑》2020年第7期44-47,共4页Information & Computer

摘　　要：笔者提出了一种图像对准估计参数的不确定性分析方法,首先把图像对准问题转化成一个残差最小值问题,然后推导出变换参数的不确定分析模型。为了量化评价图像对准精度,提出了两个具有清晰几何意义的指标。这两个指标具有几何不变性,可以准确度量图像对准误差。模拟实验数据表明:该分析模型得到的不确定性值与蒙特卡罗模拟方法计算值相一致,从而验证了本文提出图像对准估计参数不确定性分析模型的有效性。This paper proposes an uncertainty analysis approach for image registration.Firstly,we formulate the image registration problem with the defined functions,then derive the uncertainty analytical model of the transformation parameters.Then,in order to qualify the accuracy of the image registration,two indexes with clear geometric meaning are presented.The two indexes are proved that they have the frame-invariant property.Finally,simulations show that the uncertainties from the analytical model and Monte Carlo simulation are coincident within the small error in feature points,which validates the proposed model.

关 键 词：机器视觉 图像对准 不确定性 

分 类 号：TN958[电子电信—信号与信息处理]