It is noteworthy that numerous Convolution Neural Network and Transformer approaches for image super-resolution have attained significant advancements. However, these methods frequently necessitate considerable training data and computational time, which may constrain their real-world applicability and scalability. In our work, we focus on designing an efficient image super-resolution model and propose the multi-scale spatial Feature Aggregation Super-Resolution Network (FA-SRN), a novel architecture that excels at aggregating features by capitalizing on the rich, multi-layered and hierarchical nature of image data. Specifically, we incorporate the advanced spatial attention mechanism through a multi-branch architecture, which markedly amplifies the spatial representation efficacy and perceptual understanding of the images. Furthermore, we employ channel aggregation techniques to significantly enrich and refine the representation of image features, thus enhancing the model’s capability to discern subtle features within the data. Comprehensive experimental outcomes show that our proposed method attains outstanding performance with reduced parameters and lower computational demands.