Satellite cloud image super-resolution reconstruction method based on diffusion model
Menghui He, Xiangsheng Feng, Longbiao Zheng, Meiyan Wu · Journal of Physics Conference Series · 2025
Abstract Aiming at the bottleneck of satellite sensors in image resolution improvement, that is, it is difficult to output high-precision satellite cloud images, this study proposes a satellite cloud image super-resolution reconstruction method based on the diffusion model (AESR). The core architecture of the process included two parts. First, the ACNET network structure was introduced into the Efficient Prior Information Extraction Module (EPEM) to improve the efficiency of key information mining of low-resolution satellite cloud images to achieve accurate initial feature representation. The second is to construct a U-shaped network framework with an Enhanced Feature Extraction Module (EFEM) as the core, introduce the ACmix attention mechanism to strengthen cross-scale feature interaction, and extract high-order semantic information through multi-dimensional fusion. The experiment uses PSNR, SSIM, FID, LPIPS and NIQE as five indicators to evaluate, and the results show that the AESR method performs well in super-resolution tasks, which verifies its effectiveness.