The paper introduces the Lightweight Information Split Network (LISN), a novel and efficient model for single infrared image super-resolution (SR). LISN is designed to address the challenges of high computational complexity and memory demands associated with existing deep learning architectures. The model consists of four main components: shallow feature extraction, deep feature extraction, dense feature fusion, and high-resolution infrared image reconstruction. A key innovation is the Lightweight Information Split Block (LISB), which employs a sequential process to extract hierarchical features and aggregates them based on their relevance. The LISB integrates channel splitting and shift operations to balance SR performance and model complexity. Experimental results show that LISN outperforms state-of-the-art methods in terms of both SR quality and model complexity, making it suitable for resource-constrained infrared imaging applications. The code for LISN is available at <https://github.com/sad192/LISN-Infrared-Image-SR>. Infrared image SR is crucial for various vision tasks due to the unique advantages of infrared imaging systems, such as consistent performance across weather conditions and different times of the day. Enhancing the resolution of infrared images improves the extraction of fine feature information, which is essential for subsequent vision tasks. The proposed LISN combines channel splitting operations, CNNs, and shift technologies to achieve efficient SR. Key contributions include a simplified and computationally efficient shift block, a channel splitting operation to reduce computational demands, and the Lightweight Information Split Block (LISB) that integrates shift and CNN techniques. The paper reviews existing SR methods, including conventional CNNs, efficient SR models, and Transformer-based SR. It highlights the limitations of CNN-based methods in capturing global contextual information and the challenges of high computational and memory demands in efficient SR models. The introduction of Transformer-based methods and their potential in SR is also discussed. The LISN architecture is detailed, including the shallow feature extraction, deep feature extraction, dense feature fusion, and high-resolution infrared image reconstruction modules. The deep feature extraction module uses multiple LISBs to extract hierarchical features, which are then fused using a dense feature fusion module. The final high-resolution image is reconstructed by merging shallow and deep features. The model incorporates edge loss to preserve edge information, enhancing the structural fidelity of the reconstruction. Ablation studies are conducted to evaluate the effectiveness of each component of the LISN. The channel splitting operation, Residual Depth-wise Convolution Block (RDB), and Contrast-aware Channel Attention (CCA) are analyzed, showing their impact on model performance and complexity. The influence of the number of LISBs on SR performance is also examined. The performance of LISN is evaluated using the CVC-09-1K dataset, with metrics such as PSNR and SSIM. The results demonstrate that LISN achieves superior SR quality compared to other methods while maintaining a lower parameter count and computational complexity. Visual comparisons and quantitative evaluations further support the effectiveness of LISN. The paper concludes by highlighting the advantages of LISN inThe paper introduces the Lightweight Information Split Network (LISN), a novel and efficient model for single infrared image super-resolution (SR). LISN is designed to address the challenges of high computational complexity and memory demands associated with existing deep learning architectures. The model consists of four main components: shallow feature extraction, deep feature extraction, dense feature fusion, and high-resolution infrared image reconstruction. A key innovation is the Lightweight Information Split Block (LISB), which employs a sequential process to extract hierarchical features and aggregates them based on their relevance. The LISB integrates channel splitting and shift operations to balance SR performance and model complexity. Experimental results show that LISN outperforms state-of-the-art methods in terms of both SR quality and model complexity, making it suitable for resource-constrained infrared imaging applications. The code for LISN is available at <https://github.com/sad192/LISN-Infrared-Image-SR>. Infrared image SR is crucial for various vision tasks due to the unique advantages of infrared imaging systems, such as consistent performance across weather conditions and different times of the day. Enhancing the resolution of infrared images improves the extraction of fine feature information, which is essential for subsequent vision tasks. The proposed LISN combines channel splitting operations, CNNs, and shift technologies to achieve efficient SR. Key contributions include a simplified and computationally efficient shift block, a channel splitting operation to reduce computational demands, and the Lightweight Information Split Block (LISB) that integrates shift and CNN techniques. The paper reviews existing SR methods, including conventional CNNs, efficient SR models, and Transformer-based SR. It highlights the limitations of CNN-based methods in capturing global contextual information and the challenges of high computational and memory demands in efficient SR models. The introduction of Transformer-based methods and their potential in SR is also discussed. The LISN architecture is detailed, including the shallow feature extraction, deep feature extraction, dense feature fusion, and high-resolution infrared image reconstruction modules. The deep feature extraction module uses multiple LISBs to extract hierarchical features, which are then fused using a dense feature fusion module. The final high-resolution image is reconstructed by merging shallow and deep features. The model incorporates edge loss to preserve edge information, enhancing the structural fidelity of the reconstruction. Ablation studies are conducted to evaluate the effectiveness of each component of the LISN. The channel splitting operation, Residual Depth-wise Convolution Block (RDB), and Contrast-aware Channel Attention (CCA) are analyzed, showing their impact on model performance and complexity. The influence of the number of LISBs on SR performance is also examined. The performance of LISN is evaluated using the CVC-09-1K dataset, with metrics such as PSNR and SSIM. The results demonstrate that LISN achieves superior SR quality compared to other methods while maintaining a lower parameter count and computational complexity. Visual comparisons and quantitative evaluations further support the effectiveness of LISN. The paper concludes by highlighting the advantages of LISN in