A Comprehensive Review of Image Colorization Methods

Abstract

This review paper focuses on different methods that are already in use for Grayscale Image Colorization. Image Colorization can be done using various methods. In today’s world, Convolutional Neural Networks(CNNs), Autoencoders, Generative Adversarial Networks, etc are the modern techniques that are used for Image Colorization. This paper gives a comparative study of the above methodologies/architectures. Along with this, a review of different Loss functions is categorized into three categories viz. Error-based, GAN-based, Distribution-based Loss functions are described in detail. We also discuss different methods for the evaluation of an image colorizer. Finally we summarize the results of different methodologies.