基于面积投影的图像配准方法

An Image Registration Algorithm Based on Area Projection

针对低分辨率图像之间的配准精度问题,直接影响到超分辨率图像的重建质量。通常图像之间的平移和旋转,采用基于泰勒级数展开的迭代配准算法以及频域配准算法。传统的泰勒级数展开的迭代配准算法的配准精度取决于图像的低阶逼近误差及迭代过程中图像的插值近似运动变换所造成的误差。采用泰勒级数展开的配准算法进行了改进,以面积投影变换来替代原有迭代算法中的图像插值变换,这种图像变换算法更加符合图像的成像原理,仿真结果表明,算法能够有效提高低分辨率图像间平移和旋转角度的配准精度。

The image reconstruction quality is severely based on the registration precision among the low resolution images. There are two class of algorithms for translation and rotation registration ： one is based on Taylor＇s series approximation, the other utilizes the image phase relation in the frequency domain. The registration precision based on Taylor＇s series approximation mainly relies on the low -order approximation errors and the image transform errors in the iteration processing. The algorithm based on Taylor＇s series approximation is improved, which adopts the area projection instead of the interpolation for realizing the image transform. The algorithm is better accordant with the im- age formation theory. The experiments show that the algorithm improves the registration precision of translation and rotation efficiently.