结合PCM和MIE的多模态图像配准方法

针多模态图像配准问题,提出了一种基于相位一致性模型（PCM）和互信息熵（MIE）的配准方法.通过相位一致性模型同时提取多模态图像的角点和边缘特征,在边缘图上取角点的邻域,依据邻域间互信息熵的最小值在估计区域搜索匹配特征,利用RANSAC算法去除错配,进而确定待配准图像间的变换参数.实验表明：该方法达到了像素级配准精度、求解稳定,对多模态引起的非线性灰度变化、光照变化、噪声等都具有较强的鲁棒性;计算精度较基于同类特征的配准方法高,角点、边缘等几何特征综合运用在多模态图像配准中效果良好.

Method of Multimode Image Registration Based on PCM and MIE

Aimed at multimode image registration,a sort of registration method based on phase congruency model（PCM） and mutual information entropy（MIE）was proposed.By PCM the features of both corners and edges of multimode image were extracted,and by taking the corners＇ neighboring region in the image edge,the match characteristics was searched in estimation region in terms of the minimum of MIE between neighboring regions,and then the mismatching was eliminated by means of RANSAC algorithm,in order to determine the conversion parameters between images to be registered.By experiments it is shown that this registration method reaches the accuracy of pixel level,stable solution and is of better robustness to the change of nonlinear grey scale,illumination,noise and other factors.And the computation precision of this method is better than that of the registration based on congener characteristics,and the effect of applying the geometrical characteristics of corner,edge and so forth to the multimode image registration is better.