PoGaIN: Poisson-Gaussian Image Noise Modeling from Paired Samples

TL;DR

This paper introduces a cumulant-based method for accurately estimating Poisson-Gaussian noise parameters from paired noisy and noise-free images, improving over existing methods especially in real-world scenarios.

Contribution

It presents a novel cumulant-based approach leveraging paired samples for better noise modeling, filling a gap in existing Poisson-Gaussian noise estimation techniques.

Findings

Improved MSE performance over baseline methods

Robustness to outliers and image dependence

Effective in real-world noise estimation scenarios

Abstract

Image noise can often be accurately fitted to a Poisson-Gaussian distribution. However, estimating the distribution parameters from a noisy image only is a challenging task. Here, we study the case when paired noisy and noise-free samples are accessible. No method is currently available to exploit the noise-free information, which may help to achieve more accurate estimations. To fill this gap, we derive a novel, cumulant-based, approach for Poisson-Gaussian noise modeling from paired image samples. We show its improved performance over different baselines, with special emphasis on MSE, effect of outliers, image dependence, and bias. We additionally derive the log-likelihood function for further insights and discuss real-world applicability.