Deep Unfolding with Normalizing Flow Priors for Inverse Problems

TL;DR

This paper introduces a novel deep unfolding approach that integrates normalizing flow generative models into proximal gradient algorithms, explicitly modeling image priors for improved inverse problem solutions.

Contribution

It presents a new method embedding normalizing flow priors into deep unfolding algorithms, enhancing explicit prior modeling in inverse image recovery tasks.

Findings

Outperforms baseline methods in image denoising, inpainting, and deblurring.

Explicit priors improve physical plausibility and perceptual quality.

Demonstrates versatility across multiple inverse imaging problems.

Abstract

Many application domains, spanning from computational photography to medical imaging, require recovery of high-fidelity images from noisy, incomplete or partial/compressed measurements. State of the art methods for solving these inverse problems combine deep learning with iterative model-based solvers, a concept known as deep algorithm unfolding. By combining a-priori knowledge of the forward measurement model with learned (proximal) mappings based on deep networks, these methods yield solutions that are both physically feasible (data-consistent) and perceptually plausible. However, current proximal mappings only implicitly learn such image priors. In this paper, we propose to make these image priors fully explicit by embedding deep generative models in the form of normalizing flows within the unfolded proximal gradient algorithm. We demonstrate that the proposed method outperforms competitive baselines on various image recovery tasks, spanning from image denoising to inpainting and deblurring.