<p>In the field of image reconstruction, convolutional neural network (CNN)-based methods typically focus on enhancing global context modeling to address the limited long-range dependency caused by the constrained receptive field of convolutional operations. However, the reconstruction of fine image details heavily relies on high-frequency features, which have not received sufficient attention in prior studies, resulting in suboptimal outcomes, such as blurred details in reconstructed images. To address this limitation, we propose the high-frequency enhanced feature aggregation and modulation (HEFAM) module, which effectively integrates local and non-local features while enhancing high-frequency components to achieve more precise reconstruction. The HEFAM module comprises two core components: the high-frequency enhanced dual feature aggregation block (HEDFA), which simultaneously aggregates spatial and channel features while enhancing high-frequency features, and the efficient approximation of self-attention block (EASA), which models non-local features. Extensive experiments demonstrate that the proposed high-frequency enhanced feature aggregation and modulation network (HEFAMN) achieves superior performance while requiring the fewest parameters.</p>