A Comprehensive Feature Aggregation Network for Efficient Image Super-Resolution

Deep learning based models have achieved remarkable reconstruction performance in single image Super-Resolution. However, these models often involve high computational costs and substantial memory usage, which limits their practical deployment on consumer devices. To overcome these challenges, we propose a novel Comprehensive Feature Aggregation Network (CFAN) for efficient image reconstruction. Specifically, we first propose a Hybrid Pixel Attention (HPA) block to capture both local and nonlocal information at a low computational cost. To further exploit channel information, we design a Multi-Scale Channel Attention (MSCA) block to comprehensively model pixel interactions across spatial and channel dimensions. Finally, we adopt an Adaptive Feature Fusion (AFF) module, which can enhance the preservation of low-frequency structural information. Extensive experiential results demonstrate that our proposed method achieves competitive performance in terms of accuracy with fewer network parameters compared to state-of-the-art methods.