Photon Limited Non-Blind Deblurring Using Algorithm Unrolling

TL;DR

This paper introduces an end-to-end trainable unrolled neural network for photon-limited image deblurring, effectively handling shot noise and outperforming existing methods in low-light conditions.

Contribution

It proposes a novel algorithm unrolling approach with a three-operator splitting framework for photon-limited deblurring, enabling differentiable steps and end-to-end training.

Findings

Significantly improved image recovery over state-of-the-art methods

Effective handling of photon shot noise in low-light deblurring

Introduction of a new photon-limited deblurring dataset

Abstract

Image deblurring in photon-limited conditions is ubiquitous in a variety of low-light applications such as photography, microscopy, and astronomy. However, the presence of photon shot noise due to low illumination and/or short exposure makes the deblurring task substantially more challenging than the conventional deblurring problems. In this paper, we present an algorithm unrolling approach for the photon-limited deblurring problem by unrolling a Plug-and-Play algorithm for a fixed number of iterations. By introducing a three-operator splitting formation of the Plug-and-Play framework, we obtain a series of differentiable steps which allows the fixed iteration unrolled network to be trained end-to-end. The proposed algorithm demonstrates significantly better image recovery compared to existing state-of-the-art deblurring approaches. We also present a new photon-limited deblurring dataset for evaluating the performance of algorithms.