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.