针对非制冷红外焦平面探测器面阵规模较小，难以获取大尺度红外图像的问题，提出一种基于低秩矩阵恢复和邻域嵌入的单幅红外图像超分辨方法。利用低秩矩阵恢复算法学习出相似矩阵潜在的低秩分量，对恢复的低秩分量进行邻域嵌入以获得初始的超分辨估计值，再通过全局重建约束，最终获得超分辨结果。大量仿真实验结果表明，本文算法重建的图像无论是定量计算还是定性分析都获得较好的超分辨结果，该方法既保证重建的高分辨率图像均匀区域的一致性，又保留了图像的细节信息和边缘轮廓的完整性。

In this paper, we propose a single-frame infrared-image super-resolution algorithm based on low-rank matrix recovery and neighbor embedding. Low-rank matrix recovery is adopted to study the underlying structures in sub-spaces spanned by similar image patches. Specifically, training image patches are first divided into many groups and the underlying structure of each group is learned using the low-rank technique. The neighbor embedding algorithm is used on the low-rank components of low- and high-resolution image patches to produce optimal super-resolution results. Experimental results demonstrate that our proposed method can reconstruct quantitatively and perceptually high-quality images. In addition, this method not only guarantees the consistency of smooth regions in the reconstructed high-resolution image, but also retains the image details and integrity of the edge profile.