一种改进的基于互信息的自动图像配准算法  

An Improved Automatic Image Registration Algorithm Based on Mutual Information

作　　者：梅跃松[1] 杨树兴[1] 莫波[1] 

机构地区：[1]北京理工大学机电工程学院,北京100081

出　　处：《激光与红外》2007年第5期470-473,共4页Laser & Infrared

基　　金：国防基础科研项目基金(K1305060614)

摘　　要：提出了改进图像配准的基于互信息的算法,首先将参考图像和待配准图像分解到小波域并二值化,以归一化互信息作为相似性测度进行配准。配准参数搜索作用SPSA算法和鲍威尔算法相结合的优化策略。实验证明本算法可以快速、精确且不需要人工干预地实现两幅图像的配准。An improved algorithm of image registration based on mutual information was proposed, the reference image and registered image were decomposed into wavelet field and binaried. The normalized mutual information was used to be a scale of similarity to two registered images. SPSA algorithm and Powell algorithm were combined to optimize the parameter searching process. Experiments show that this algorithm can register two images rapidly, accurately and automatically.

关 键 词：图像配准 互信息 小波分解 SPSA算法 鲍威尔法 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]