High-resolution remote sensing imagery is essential for critical applications such as precision agriculture, urban management planning, and military reconnaissance. Although significant progress has been made in single-image super-resolution (SISR) using generative adversarial networks (GANs), existing approaches still face challenges in recovering high-frequency details, effectively utilizing features, maintaining structural integrity, and ensuring training stability—particularly when dealing with the complex textures characteristic of remote sensing imagery. To address these limitations, this paper proposes the Improved Residual Module and Attention Mechanism Network (IRMANet), a novel architecture specifically designed for remote sensing image reconstruction. IRMANet builds upon the Super-Resolution Generative Adversarial Network (SRGAN) framework and introduces several key innovations. First, the Enhanced Residual Unit (ERU) enhances feature reuse and stabilizes training through deep residual connections. Second, the Self-Attention Residual Block (SARB) incorporates a self-attention mechanism into the Improved Residual Module (IRM) to effectively model long-range dependencies and automatically emphasize salient features. Additionally, the IRM adopts a multi-scale feature fusion strategy to facilitate synergistic interactions between local detail and global semantic information. The effectiveness of each component is validated through ablation studies, while comprehensive comparative experiments on standard remote sensing datasets demonstrate that IRMANet significantly outperforms both the baseline and state-of-the-art methods in terms of perceptual quality and quantitative metrics. Specifically, compared to the baseline model, at a magnification factor of 2, IRMANet achieves an improvement of 0.24 dB in peak signal-to-noise ratio (PSNR) and 0.54 in structural similarity index (SSIM); at a magnification factor of 4, it achieves gains of 0.22 dB in PSNR and 0.51 in SSIM. These results confirm that the proposed method effectively enhances detail representation and structural reconstruction accuracy in complex remote sensing scenarios, offering robust technical support for high-precision detection and identification of both military and civilian aircraft.

Infrared imaging technology plays an indispensable role in critical applications such as military surveillance, autonomous driving, and medical diagnostics. However, its inherent low-resolution and low-contrast characteristics often limit operational performance. While deep learning-based super-resolution (SR) techniques offer a software-driven solution, models face severe feature redundancy caused by simply stacking deep layers, and a lack of discriminative power in distinguishing critical textures from thermal noise. To address these issues, we proposed a novel Convolutional and Attention-based Super-Resolution Network (CASRNet). The novelty of our model lies in the synergistic fusion of a channel splitting (CS) strategy and a dual attention mechanism. First, the CS strategy decomposes feature maps into parallel streams, extracting diverse and less redundant representations. Second, a novel channel and spatial attention residual block (CSA_ResBlock) was designed to adaptively focus on informative feature channels and critical spatial boundaries. Quantitatively, CASRNet achieved superior performance on public benchmarks. Specifically, for the FLIR dataset (× 2), our model achieved a peak signal-to-noise ratio (PSNR) of 39.73 dB and structural similarity (SSIM) of 0.9639, outperforming the state-of-the-art infrared-specific model TherISuRNet by 0.48 dB and standard models like VDSR by 0.15 dB. Similar robust improvements (e.g., an exceptional 40.45 dB PSNR on the ThermalTau2 dataset) demonstrated the general applicability and high fidelity of CASRNet for real-world infrared enhancement tasks.