<p>Stereo image super-resolution (StISR) presents unique challenges compared to single image super-resolution (SISR), not merely due to the presence of two input images but primarily because of the need to effectively leverage stereo correspondence while preserving geometric consistency. The task requires addressing disparity estimation, cross-view feature fusion, and depth-aware reconstruction, making it significantly more complex. With recent advancements in deep learning, various models have been proposed to enhance the quality of super-resolved stereo images. StISR has wide-ranging applications, including autonomous vehicles, virtual reality, augmented reality, and medical imaging. In this paper, we provide a comprehensive review of StISR, summarizing recent advancements and categorizing existing models based on four key criteria: network design, loss function, framework, and training datasets. To better analyze network structures, we classify StISR models into four primary categories–attention-based models, Transformer-based models, perceptual-based models, and other architectures. Additionally, we present an extensive comparative analysis of these models, evaluating their performance across different benchmarks. Finally, we discuss the major challenges in StISR, such as stereo alignment, depth consistency, and computational efficiency, while outlining potential future research directions to further advance the field.</p>