摘要： 针对同时遭受离焦模糊退化和几何形变的图像，提出一种基于Legendre正交矩不变量的图像配准方法.采用Harris特征点探测算法提取图像特征点，然后构造Legendre正交矩的模糊和几何混合不变量作为衡量尺度以获得特征点的对应关系，通过该对应关系估计形变参数完成配准过程. 实验结果表明该方法能有效解决该类形变图像的配准问题，与其他方法相比能获得更准确的结果.

Abstract: We propose a new Legendre moment invariant based method to deal with registration of images which are simultaneously degraded by out-of-focus blurring and geometric deformation. We use the Harris corner detector to extract feature points, and construct a new set of combined blur and geometric invariantsof Legendre moment as a measure to establish correspondence between feature points of the images to be registered. This correspondence is then used to estimate the transformation parameters to realize image registration. Experimental results show that the proposed method can effectively register degraded images with better performance compared to some other methods.