Abstract: Sub-pixel image registration is an essential part of super-resolution image reconstruction algorithms. Unfortunately, when the rotation angel between input images is small, the conventional frequency domain algorithms cannot obtain an accurate estimation. Thus, an improved approach is proposed by taking into account both the high frequency and low frequency coefficients. The numerical results show that the proposed algorithm has a better performance in estimating the small rotation angles.

Key words: super resolution, frequency domain, rotation estimation, spectrogram

摘要： 亚像素级配准是超分辨率重构的前提，其中旋转角估计的准确程度对重构效果有直接影响。针对传统的频域内旋转角估计在小角度估计不精确的缺点，提出了用部分幅值进行配准的改进算法。实验结果表明，改进算法对小角度旋转角估计比传统算法效果好。

关键词: 超分辨率重构, 频域, 旋转角估计, 频谱图