As low-resolution (LR) images often miss a large amount of feature information, the super-resolution (SR) algorithm that directly maps and reconstructs a single LR infrared image faces a considerable challenge due to the lack of prior features. We propose an infrared and visible light dual-camera SR imaging scheme based on a visible light texture transfer (VTTISR) method, which fills in the missing feature information of LR infrared images by transforming visible light texture features. The proposed VTTISR method consists of a multi-scale texture progressive transfer network and a cross-scale residual aggregation reconstruction network. The multi-scale texture progressive transfer network migrates visible light texture features to infrared features to generate multi-scale texture transfer features. The cross-scale residual aggregation reconstruction network sequentially fuses texture features at corresponding scales and generates 4-magnification SR infrared image. Experimental results show that VTTISR has made significant progress in quantitative image evaluation with higher efficiency.

由于低分辨率（LR）图像经常遗漏大量特征信息，直接映射和重建单个LR红外图像的超分辨率（SR）算法由于缺乏先验特征而面临相当大的挑战。我们提出了一种基于可见光纹理转移（VTTISR）方法的红外和可见光双摄像头SR成像方案，该方案通过变换可见光纹理特征来填补LR红外图像缺失的特征信息。所提出的 VTTISR 方法由多尺度纹理渐进传输网络和跨尺度残差聚合重建网络组成。多尺度纹理渐进传递网络将可见光纹理特征迁移到红外特征，生成多尺度纹理传递特征。跨尺度残差聚合重建网络依次融合相应尺度的纹理特征，生成 4 倍 SR 红外图像。实验结果表明，VTTISR在量化图像评价方面取得了显着进展，效率更高。