一种混合红外云图超分辨率重建算法

红外云图具有分辨率较低、图幅大和纹理丰富等特点。针对相关研究目前在算法效率优化和局部细节分析方面仍有不足，提出了一种混合超分辨率重建算法。该方法结合双三次插值重建方法和基于稀疏表达的重建方法在不同类型图像区域中的各自优势，利用方差将滑动窗口中的图像块区分为平坦和边缘两种类型；采用双三次插值方法重建平坦型图像块，采用基于稀疏表达的方法重建边缘型图像块。利用目视、峰值信噪比(Peak Signal-to-Noise Ratio, PSNR)以及残差图评估了算法效果。实验结果表明，本文方法在PSNR指标上比插值法平均提高了1 dB，比稀疏法也略有提升；经局部观察发现，改进重建结果中平坦区域噪声减少；该方法的重建耗时明显减少。

A Super-Resolution Reconstruction Algorithm for Hybrid Infrared Cloud Images

Infrared cloud images have the features of low resolution, large size and rich texture. The current research on super-resolution reconstruction is still inefficient in the optimization of algorithm efficiency and the analysis of local details. To deal with these problems, a hybrid super-resolution reconstruction algorithm is proposed and the effect of the algorithm is evaluated by PSNR combined with residual graph. This method combines the advantages of the bicubic interpolation method and the sparse representation-based method in the reconstruction of different types of images. Through variance, the image blocks in the sliding window are divided into two types of flat and edge. The bicubic interpolation method is used to reconstruct the flat image block, and the edge block is reconstructed by the sparse representation method. The experimental results show that the PSNR index of the hybrid method is 1 dB higher than that of the interpolation method and is slightly higher than that of the sparse method. Moreover, it is found in the partial observation that the noise is reduced in the plain region after reconstruction and the reconstruction time is significantly.