Image denoising: from traditional methods to neural networks

Image denoising is one of the most widely explored problems in computational imaging. Over the years, a rich variety of methods have been proposed to deal with this issue with inspiration coming from multiple fields. Until recently, “traditional methods” exploiting the self-similarity assumption achieved state-of-the-art results. In the last five years however, the development of deep learning has revolutionized computer vision, through significant accuracy improvements, denoising task being no exception. The goal of this talk is to give an overview of both traditional and deep-learning based methods for image denoising and to understand how they can reinforce each other.