图像分析应用对图像配准的精度和速度要求越来越高,传统配准方法很难满足要求.为此,提出一种基于二人博弈的配准方法,该方法将两种基于不同图像特征(如特征点和轮廓)的配准方法视为两个局中人,通过寻求二人博弈的均衡点配准图像;配准过程中,两个局中人通过信息共享改善配准精度和速度.新方法很好地结合了不同图像特征的优点,克服了采用单一特征进行配准的不足.针对医学、遥感和航拍图像的实验表明,在选择合适图像特征的前提下,新方法的配准速度和精度都优于基于SIFT的配准方法.%Image analyze field requires higher accuracy and more efficient image registration approaches, contrast to traditional approaches.This paper proposed an image registration method based on two-player game theory, the novel method viewed two feature( such as feature points and contours) based image registration methods as two players,achieved the registration via finding a Nash equilibrium of the two-player game.By this way, the accuracy and efficiency of the novel method could be improved by sharing players' information during the registration process.The novel algorithm combined the virtues of different features,and eliminated the defects of registering images by using only one image feature.Experimental results on medical images, remote sensing images and aerial images show that, when adopt appropriate features, the proposed approach can register images with higher accuracy and higher efficiency contrast to the methods based on SIFT.
展开▼
