Colorization of grayscale images is a meaningful computer vision task, falling within a wide spectrum of applications—from historical photographs restoration to medical imaging analysis. The recent breakthroughs in the field of deep learning have given this field a significant turn and have opened ways for automatic realistic colorization with very limited human input. This review systematically evaluates several deep learning approaches, including CNNs, GANs, and transformers, by examining their methodologies and performances, along with their limitations. It provides an overall review of current trends and future prospects of image colorization using deep learning and highlights challenges in its way like color ambiguity, limitation of datasets, and computational complexity.