Image super-resolution using deep learning is a prominent research area in computer vision that focuses on reconstructing high-resolution images from low-resolution inputs, enhancing image quality and preserving fine details. Early approaches utilized convolutional neural networks (CNNs) such as SRCNN to learn end-to-end mappings between low- and high-resolution images. Subsequent research introduced deeper and more sophisticated architectures including VDSR, EDSR, and residual networks, as well as generative adversarial networks (GANs) like SRGAN for producing perceptually realistic high-frequency details. Recent advances leverage attention mechanisms, transformer-based architectures, and multi-scale or progressive learning strategies to improve reconstruction quality, robustness, and computational efficiency. Applications span medical imaging, satellite and aerial imagery, surveillance, video enhancement, and consumer photography. Current studies also explore lightweight and real-time models for edge deployment, unsupervised or self-supervised super-resolution, and integration with multimodal data, establishing deep learning-based super-resolution as a key technology for high-fidelity image enhancement.