一种基于新的相似性测度的自动图像配准算法.pdf.pdf

第 28 卷第 4 期增刊仪器仪表学报 Vol. 28 No . 4
2007 年 4 月 Chinese Jour nal of Scientif ic Instrument Apr. 2007
一种基于新的相似性测度的自动图像配准算法 3
梅跃松, 杨树兴, 莫波
(北京理工大学机电工程学院北京 100081)
摘要: 图像配准是图像处理的一个重要技术,被广泛地用于遥感图像、医学影像、视频监控等诸多领域中。本文提出了一种基
于新的相似性测度的配准算法,首先分析 2 幅图像的联合直方图点集的分布情况,基于此定义了直方图点集的散度公式,并将其
作为相似性测量。为加速参数的搜索过程,配准是在小波域内进行的,并使用 Powell 算法来优化搜索策略。实验证明,相对于基
于互信息的图像配准算法,本算法参数优化方法选择可以更灵活,时间消耗更少,噪声鲁棒性更优。
关键词: 图像配准; 直方图散度; 互信息
Automat ic ima ge registrat ion algor ithm ba sed on a novel simila rity metr ic
Mei Yuesong , Yang Shuxing , Mo Bo
( School of Mcchat ronic E ngineer ing , Bei ji ng i nstitute of technology , Bei jing 100081 , Chi na)
Abstract : Image regist ration is a kind of impo rtant technology of i mage p rocessi ng , and is u sed in the fiel d of remot e
sensing , medical image , vi deo surveillance , and so on . An image regist ratio n al gorit h m based o n a novel similarity
measure was p ropo sed , and t he point set di stri butio n of joint histogram was analyzed. Its divergence fo rmula was de2
fined and used as t he si mil arit y metric. To speed up searching t he regi st ration p arameters , all were done in the wavelet
field and Powell method was used to opti mize t his p roces s. Exp eri ment s sho w t hat t hi s al gorith m can apply wider opti2
mizat io n met ho ds an d save mo re time and have better anti2noise p p ared wit h t he al gorit h m based o