一种有效的图像超分辨率重建算法

Effective super-resolution image reconstruction algorithm

本文提出了一种基于频域图像配准和边缘指导插值的图像超分辨率重建算法。该算法先在频域中对低分辨率(LR)混迭图像序列进行精确配准,再利用边缘指导插值算法对配准后的图像进行插值,实现高分辨率(HR)图像重建。考虑到信号低频部分的信噪比最高又不存在混迭,图像配准中只使用序列的低频信息进行运算。插值算法则是先计算LR图像各像素点的局部协方差系数,再利用LR和HR图像协方差间的几何对偶性来计算HR图像插入像素点的值。实验结果表明,本文提出的算法能有效地实现序列超分辨率重建。

A super-resolution image reconstruction algorithm based on frequency domain image registration and edge-directed interpolation was proposed. Firstly,a set of aliased low-resolution images were registered precisely through a frequency domain technique. A high-resolution image was then reconstructed using a new edge-directed interpolation. The registration algorithm only used low frequency information, because this was the aliasing-free part of the signals with the highest signal-to-noise ratio. The basic idea of interpolation was to first estimate local covariance coefficients from a low-resolution image, and then used these covariance estimates to adapt the interpolation at a higher resolution based on the geometric duality between the low-resolution covariance and the high-resolution covariance. The experiment results show that the algorithms can reconstruct super-resolution image effectively.