<p>Acquiring and transmitting high-resolution images typically demands high power consumption and larger physical space requirements, significantly affecting the practicality and reliability of devices. Deploying single image super-resolution (SISR) models based on convolutional neural networks in devices provides a feasible solution to enable low-power terminals to obtain high-resolution images. Existing methods often significantly increase network complexity to enhance performance or excessively sacrifice effectiveness for lightweight design. To address this issue, we propose a novel SISR network called Progressive Distillation Super-Resolution Network (PDSRN), which achieves better performance without significantly increasing complexity. Built upon stacked residual dense blocks for comprehensive feature extraction, the proposed model incorporates a compact and efficient information distillation block at both local and global levels, progressively distilling features at each stage. The features extracted from the basic blocks are individually processed along distinct paths by the local distillation block (LDB). By stacking residual dense information distillation groups, each local feature is fused at the end of the network and processed by a global distillation block (GDB). Since LDB and GDB utilize only a small number of features and convolution kernels, their inclusion minimally increases the overall parameter count of the SISR network. Experimental results demonstrate that the proposed method achieves competitive performance in both metrics and visual quality, delivering an average improvement of 0.1&#xa0;dB across all benchmark datasets and scaling factors while incurring only a minimal increase in network complexity. Furthermore, the method achieves performance improvements on real-world datasets, demonstrating its strong generalization capability.</p>