摘要
基于互信息的图像配准方法,已经广泛应用于图像配准领域。但互信息图像配准方法容易受到局部极值的干扰,难以得到最优解。对互信息图像配准中互信息的表征、图像插值方法以及优化搜索算法三个要素做了探讨,尤其针对常用的Powell搜索算法的不足,提出了基于互信息和二级搜索的图像配准算法。该算法以标准互信息为图像相似测度,利用PV插值法平滑搜索空间,采用Simplex算法进行一级粗配准,采用Powell算法进行二级精配准。仿真结果表明,二级搜索配准算法能够有效地克服局部极值,提高计算速度,用于大差异图像配准。
Mutual information-based image registration has been widely applied to the field of image registration. But this method is easy to be interfered by local extrema that can＇t get the optimal solution. Discussions are made above the measure of mutual information, image interpolation and optimizer, which are the three crucial factors of mutual information based image registration. A two stage search algorithm as optimizer is proposed, especially considering the drawback of common used Powell＇s direction set method. In this algorithm, normal mutual information is selected as similarity measure, partial volumn interpolation is used to smooth search space, Simplex method is used for the first stage coarse registration and Powell method for the second stage fine registration. Experimental results show that two stage search algorithm can effectively restrain local extrema, improve calculation speed, and be applied to great difference image registration.
出处
《微型机与应用》
2013年第9期1-4,共4页
Microcomputer & Its Applications
