The paper introduces a deep learning method for single image super-resolution (SR) using a deep convolutional neural network (CNN). The proposed method, named Super-Resolution Convolutional Neural Network (SRCNN), learns an end-to-end mapping between low- and high-resolution images, with minimal pre/post-processing. SRCNN is designed to be lightweight yet achieve state-of-the-art restoration quality and fast speed for practical online usage. The authors show that traditional sparse-coding-based SR methods can be viewed as a deep CNN, but SRCNN jointly optimizes all layers, including patch extraction, non-linear mapping, and reconstruction. Experiments demonstrate that SRCNN outperforms existing methods in terms of PSNR and speed, and can be further improved with larger datasets and models. The paper also explores the learned filters and the impact of filter size and number on performance.The paper introduces a deep learning method for single image super-resolution (SR) using a deep convolutional neural network (CNN). The proposed method, named Super-Resolution Convolutional Neural Network (SRCNN), learns an end-to-end mapping between low- and high-resolution images, with minimal pre/post-processing. SRCNN is designed to be lightweight yet achieve state-of-the-art restoration quality and fast speed for practical online usage. The authors show that traditional sparse-coding-based SR methods can be viewed as a deep CNN, but SRCNN jointly optimizes all layers, including patch extraction, non-linear mapping, and reconstruction. Experiments demonstrate that SRCNN outperforms existing methods in terms of PSNR and speed, and can be further improved with larger datasets and models. The paper also explores the learned filters and the impact of filter size and number on performance.