摘要： 假设图像频谱是有限波段的，将低分辨率图像混迭的离散傅里叶变换系数与未知场景连续傅里叶变换的相应采样点相联系。利用矩阵相乘描述成像模型中各个元素之间的关系，基于矩阵秩的关系构造目标函数。通过对目标函数进行最小化，可以得到正确的序列图像相对位置关系和连续傅里叶变换的系数，将高精度配准与后期图像重建相结合。实验结果证明，该方法取得了良好效果。

关键词: 超分辨率, 图像重建, 傅里叶变换

Abstract: This paper supposes the band of image frequency is limited and makes the Discrete Fourier Transform(DFT) coefficients of low-resolution image aliasing and the Continuous Fourier Transform(CFT) of the corresponding sampling point in unknown continuous scene relevant. It uses matrix multiplication to describe element of imaging model and constructs an object function based on the relationship of rank of matrix of image model. By minimizing the object function, the accurate relative location relation of sequence image CFT coefficients can be got. The reconstruction and registration are combined with together. Experimental results confirm that this algorithm works well.

Key words: super-resolution, image reconstruction, Fourier transformation

中图分类号: 

• TN911.73