Infrared imaging is widely used in remote sensing and military target recognition due to its strong resistance to interference in complex environments. However, imaging mechanisms and hardware limitations cause infrared images to have low-resolution, sparse textures, and significant background noise, which severely restrict the detection of small targets such as low-altitude drones and weak thermal emitters. To overcome these limitations, we propose a lightweight Local–Global Dual-Path Feature Fusion Network (LDFF-Net) that enhances the resolution and quality of infrared images, providing high-quality inputs for subsequent detection tasks. The network includes a Small Target Feature Recognition Module (STFRM) composed of three key components. The Enhanced High Frequency Perception Module (EHFPM) strengthens high-frequency details of small targets while suppressing noise, enabling robust local feature extraction. The State-Space Model (SSM) captures long-range dependencies with linear complexity and models semantic relationships between targets and background to compensate for the limited receptive field of local features. The Adaptive Feature Fusion Unit (AFFU) combines local and global features adaptively to improve the saliency of small targets. During training, we introduce a realistic degradation process based on visible-light images to generate training samples that include complex degradation patterns and noise, which enhances the model’s robustness and generalization. Evaluation on the ARCHIVE and SIRST datasets demonstrates that LDFF-Net outperforms existing state-of-the-art methods across eight widely used full-reference and no-reference metrics, including PSNR, LPIPS, FID, and NIQE. This result confirms the model’s effectiveness in enhancing both the super-resolution and detection performance of infrared small target images. The code and pretrained model weights are publicly available at https://github.com/98Hao/LDFF-Net.

Intrinsic decomposition separates an image into reflectance and shading, which contributes to image editing, augmented reality, etc. Despite recent efforts dedicated to this field, effectively separating colored illumination from reflectance and correctly restoring it into shading remains an challenge. We propose a deep intrinsic decomposition method to address this issue. Specifically, by transforming intrinsic decomposition process in RGB image domains into the combination of intensity and chromaticity domains, we propose a novel macro intrinsic decomposition network framework. This framework enables the generation of finer intrinsic components through more relevant features propagation and more detailed sub-constraints guidance. In order to expand the macro network, we integrate multiple attention mechanism modules in key positions of encoders, which enhances the extraction of distinct features. We also propose a skip connection module based on specific deep features guidance, which can filter out features that are physically irrelevant to each intrinsic component. Our method not only outperforms state-of-the-art methods across multiple datasets, but also robustly separates illumination from reflectance and restores it into shading in various types of images. By leveraging our intrinsic images, we achieve visually superior image editing effects compared to other methods, while also being able to manipulate the inherent lighting of the original scene.