Learning Diverse Image Colorization
This addresses the need for diverse colorization outputs in image processing, though it is incremental as it builds on prior VAE and GAN methods.
The paper tackled the problem of ambiguous image colorization by modeling its intrinsic diversity to produce multiple spatially-coordinated colorizations, demonstrating better performance than existing CVAE and cGAN models.
Colorization is an ambiguous problem, with multiple viable colorizations for a single grey-level image. However, previous methods only produce the single most probable colorization. Our goal is to model the diversity intrinsic to the problem of colorization and produce multiple colorizations that display long-scale spatial co-ordination. We learn a low dimensional embedding of color fields using a variational autoencoder (VAE). We construct loss terms for the VAE decoder that avoid blurry outputs and take into account the uneven distribution of pixel colors. Finally, we build a conditional model for the multi-modal distribution between grey-level image and the color field embeddings. Samples from this conditional model result in diverse colorization. We demonstrate that our method obtains better diverse colorizations than a standard conditional variational autoencoder (CVAE) model, as well as a recently proposed conditional generative adversarial network (cGAN).