大倍数超分辨率图像重构的通用快速算法

A General Fast Algorithm for Large-magnification Super-resolution Image Reconstruction

为了解决超分辨率图像重构的计算复杂性随着放大倍数的加大而呈指数规律急剧上升这一问题 ,文中提出了一个通用的快速超高分辨率图像重构算法 .该算法利用各低分辨率图像之间的位移关系将所有低分辨率图像进行分组 ,然后分别对每个组进行超分辨图像重构 ,并重复此分组重构过程直至获得预期的分辨率 .实验结果表明 ,该快速算法能够在保持重构效果基本不变的同时 ,较大地降低超分辨率图像重构的计算复杂性 .

In order to tackle the difficult problem that the computational complexity of super-resolution image reconstruction rapidly and exponentially increases with the increase in amplification, this paper proposes a general fast super-resolution image reconstruction algorithm, in which the translation relation of multi low-resolution images is used to divide the images into several groups. Each group is then reconstructed by means of the super-resolution algorithm, and the grouping-reconstructing process is continuously repeated till the desired resolution is achieved . Experimental results show a significant reduction of computational complexity of the algorithm while a basically constant visual output quality is retained.