Bayesian Image Super-Resolution with Deep Modeling of Image Statistics

TL;DR

This paper introduces a Bayesian framework for image super-resolution that models natural image statistics with deep learning, demonstrating superior generalization and effectiveness in unsupervised settings across various restoration tasks.

Contribution

It proposes a novel Bayesian image restoration method combining smoothness and sparsity priors, with a deep neural network implementation and unsupervised training strategy for super-resolution.

Findings

Outperforms existing methods in diverse SISR tasks

Shows robustness to noise and kernel variations

Effective in unsupervised learning scenarios

Abstract

Modeling statistics of image priors is useful for image super-resolution, but little attention has been paid from the massive works of deep learning-based methods. In this work, we propose a Bayesian image restoration framework, where natural image statistics are modeled with the combination of smoothness and sparsity priors. Concretely, firstly we consider an ideal image as the sum of a smoothness component and a sparsity residual, and model real image degradation including blurring, downscaling, and noise corruption. Then, we develop a variational Bayesian approach to infer their posteriors. Finally, we implement the variational approach for single image super-resolution (SISR) using deep neural networks, and propose an unsupervised training strategy. The experiments on three image restoration tasks, \textit{i.e.,} ideal SISR, realistic SISR, and real-world SISR, demonstrate that our method has superior model generalizability against varying noise levels and degradation kernels and is effective in unsupervised SISR. The code and resulting models are released via \url{https://zmiclab.github.io/projects.html}.