Image colorization of historical moments provides an immersive experience with emotional impact by restoring it with plausible colors and enhancing the quality of the image. The method of image colorization has improved over time. Unlike previous studies, which heavily relied on manual adjustment and user interaction, recent research automatized the process with deep learning techniques. It enabled the model to colorize grayscale photos, recognizing their semantic colors. This paper investigates different automatic image colorization methods, including vanilla convolutional neural networks (CNN), autoencoders, generative adversarial networks (GANs), transformer-based models, etc. We examine the development of image colorization, evaluating each model qualitatively by directly comparing the result photos and quantitatively with the final loss value.