基于Legendre正交矩的模糊形变图像的配准方法  

A Registration Method for Blur Deformation Image Based on Orthogonal Legendre Moment

作　　者：左欣[1,2] 戴修斌[1,3] 张辉[1] 罗立民[1] 舒华忠[1] 

机构地区：[1]东南大学计算机科学与工程学院影像科学与技术实验室,江苏南京210096 [2]江苏科技大学计算机科学与工程学院,江苏镇江212003 [3]南京邮电大学地理与生物信息学院,江苏南京210003

出　　处：《电子学报》2011年第12期2824-2830,共7页Acta Electronica Sinica

基　　金：国家973重点基础研究发展规划(No.2010CB732503);国家自然科学基金(No.60873048);江苏省自然科学基金(No.SBKS200910055;No.BK2008279)

摘　　要：针对同时含有模糊和几何形变的图像,本文提出一种新的基于Legendre正交矩模糊和几何混合不变量的图像配准方法.该方法首先利用Harri-sLaplace算子检测出图像的特征点,然后构造Legendre矩混合不变量,并将其作为特征点的描述子获取特征点的对应关系,接着通过该对应关系估计图像间的形变参数,最后利用插值方法实现图像的配准.实验结果表明:本文方法能有效地解决含有混合形变的图像配准问题,并且和其他配准方法相比能获得更加准确的结果.To register the images which are simultaneously distorted by blur degradation and geometric transformation,we propose a new registration method based on the mixture of blurring and geometric invariant and of the orthogonal Legendre moment.It constructs a new set of combined blur and geometric invariants of Legendre moment as feature descriptor to establish the correspondence of feature points between images which are extracted by Harris-Laplace detector.Then,with the help of the correspondence above,transformation parameters are estimated and finally the distorted images can be aligned through interpolation.The experimental results show that our method can efficiently register the distorted images with better performance than the other methods in the literature.

关 键 词：图像配准 Harris-Laplace探测器 LEGENDRE正交矩 混合不变量 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]