中文摘要：

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.

英文摘要：

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.