We propose a novel approach to automatically produce multiple colorized versions of a grayscale image. Our method results from the observation that the task of automated colorization is relatively easy given a low-resolution version of the color image. We first train a conditional PixelCNN to generate a low resolution color for a given grayscale image. Then, given the generated low-resolution color image and the original grayscale image as inputs, we train a second CNN to generate a high-resolution colorization of an image. We demonstrate that our approach produces more diverse and plausible colorizations than existing methods, as judged by human raters in a "Visual Turing Test".
@article{arxiv.1705.07208,
title = {PixColor: Pixel Recursive Colorization},
author = {Sergio Guadarrama and Ryan Dahl and David Bieber and Mohammad Norouzi and Jonathon Shlens and Kevin Murphy},
journal= {arXiv preprint arXiv:1705.07208},
year = {2017}
}