A Bayesian image inpainting method with additive Gaussian process

Abstract

This paper investigates the application of Gaussian process in image inpainting, which uses the remain region in the damaged image to train the Gaussian process model, and then makes prediction for the missing parts. Additive high order kernels are employed to describe the input interaction. This additional structure of Gaussian process can improve the interpretability of the model and increase its predictive power. Experimental results in image inpainting show the additive Gaussian process leads better prediction.