Image Denoising and Colorization Based on Plug and Play Framework

Abstract

A digital image has become an essential carrier of information transmission, but it will inevitably be destroyed during the transmission process. For image denoising or image colorization, there are already many excellent algorithms, and there have been promising developments in the field of deep learning. However, in these classic methods, few algorithms can accomplish image denoising and colorization simultaneously. This paper proposes a plug and play framework to solve this problem, which combines the advantages of generative adversarial networks and plug and play framework. Moreover, we design a lightweight conditional GAN to achieve the purpose of training with fewer computing resources. Experiments on CFP and Landscape datasets show that this method can effectively complete image denoising and image colorization tasks.