基于对齐度和互信息的红外与可见光图像配准  

Infrared and visible image registration based on alignment and mutual information

作　　者：张家良[1] 金鑫[1] 赫雷[1] 周克栋[1] 

机构地区：[1]南京理工大学,江苏南京210094

出　　处：《现代电子技术》2013年第12期5-8,共4页Modern Electronics Technique

摘　　要：针对红外与可见光图像的特点,在此提出了一种结合边缘对齐度与互信息的图像配准方法。首先通过小波变换边缘检测得到红外与可见光图像的边缘图像,并将对齐度和归一化互信息自适应加权平均得到新的相似性测度函数,最终通过计算相似性测度函数的极值求得待配准图像间的变换参数。实验结果表明,该方法可减少配准所需的时间,具有更高的精确性和鲁棒性。According to the characteristics of infrared image and visible image, a registration method for infrared and visi- ble images is proposed in this paper, in which marginal alignment and mutual information are combined. The edge images of the infrared and visible images are achieved by wavelet transform. The alignment metric and normalized mutual information are cal- culated to obtain new similarity measuring function. The transform parameters between images under registration are obtained according to the calculated result of the extreme value of the similarity measuring function. Experiment results show that this method can reduce the time needed for registration, and has higher accuracy and robustness.

关 键 词：图像配准 红外图像 可见光图像 归一化互信息 

分 类 号：TN919.34[电子电信—通信与信息系统]