In recent years, with the rapid development of deep learning, convolutional neural network-based approaches for image super-resolution (SR) have made great progress. However, the majority of these methods have high computational complexity and substantial memory demands due to their deep and intricate network designs, which pose significant challenges for deployment on resource-constrained devices. In this paper, we propose a lightweight hybrid domain enhancement network (HDEN). As the core of HDEN, a hybrid domain enhancement module employs two parallel branches dedicated to feature enhancement in both the spatial and frequency domains, thereby improving feature representation. Specifically, a spatial domain enhancement block (SDEB) is introduced to extract multi-scale features, and a frequency domain enhancement block (FDEB) is proposed to explore frequency-related information. SDEB comprises four parallel branches: three branches leverage wide-activated residual units with varying dilation factors to expand the receptive field while the remaining branch retains the original scale information through a direct connection. Recognizing the critical role of frequency domain features in image SR, FDEB applies a wavelet transform to shift features from the spatial domain to the frequency domain, enabling the effective utilization of high-frequency sub-bands to enhance sharp details, such as edges, in the feature representation. Experimental results demonstrate that the proposed HDEN outperforms compared to lightweight methods in both quantitative metrics and visual quality.