均方差图象匹配系统的最佳窗口函数

OPTIMUM WINDOW FUNCTIONS FOR THE MEAN SQUARD DIFFERENCE IMAGE MATCfflNG SYSTEM

本文假定图象信号和附加噪声是零均值齐次高斯随机场,且二次可微,实测图象相对于参考图象有小的几何失真,导出了描述匹配系统配准特性的两个参数(极大极小比率和定位精度)的解析表达式,以此为优化目标,用变分学方法导出了均方差图象匹配系统最佳窗口函数Ⅰ、Ⅱ的实域和频域表达式,最后用计算机模拟验证了最佳窗口函数对图象匹配系统定位特性的改善。

This paper supposes that the image signal and additional noise are independent homogeneous Gaussian random field with zero mean, both of them are 2-order differentible, and the sensor image and reference image differ by small geometric distortion. The analytical expressions of two parameter (i.e. Max-Min ratio and local accuracy) which discribe the acquisition performance and matching accuracy respectively are derived. The two parameters are used as the objects to be optimized. The expressions of the optimum window function Ⅰ, Ⅱ for the mean squard difference image matching system are derived in frequency domain and real domain. Computer simulation results are used to test the advantage of the optimum window functions for matching performance.