Greyscale image colorization remains a challenging task due to its subjective nature and the complexity of capturing color semantics accurately. In this work, a strategy to enhance greyscale images using a hybrid ResNet and U-Net architecture, leveraging the comprehensive COCO dataset for training and validation. Our methodology involves employing ResNet for initial feature extraction, which captures high-level semantic information, followed by U-Net for refined colorization, enabling precise localization of color details. Our approach exhibits robustness in handling diverse image characteristics and scenarios, making it suitable for real-world applications such as image restoration, digital content creation, and medical imaging. It is also contributing to advancements in computer vision research and practical image enhancement techniques.