RENOIR - A Dataset for Real Low-Light Image Noise Reduction

TL;DR

This paper introduces a real low-light image noise dataset, proposes a noise estimation method, and evaluates several denoising algorithms, revealing differences in performance on real versus synthetic noise.

Contribution

The paper provides a new dataset of real low-light noisy images with aligned low-noise counterparts and introduces a noise estimation technique for accurate noise level assessment.

Findings

Multi-Layer Perceptron, Bilevel-MRF, and NL-means outperform BM3D on synthetic noise

Denoising algorithms perform worse on real low-light noise than on artificial noise

The dataset enables more realistic evaluation of denoising methods

Abstract

Image denoising algorithms are evaluated using images corrupted by artificial noise, which may lead to incorrect conclusions about their performances on real noise. In this paper we introduce a dataset of color images corrupted by natural noise due to low-light conditions, together with spatially and intensity-aligned low noise images of the same scenes. We also introduce a method for estimating the true noise level in our images, since even the low noise images contain small amounts of noise. We evaluate the accuracy of our noise estimation method on real and artificial noise, and investigate the Poisson-Gaussian noise model. Finally, we use our dataset to evaluate six denoising algorithms: Active Random Field, BM3D, Bilevel-MRF, Multi-Layer Perceptron, and two versions of NL-means. We show that while the Multi-Layer Perceptron, Bilevel-MRF, and NL-means with soft threshold outperform BM3D on gray images with synthetic noise, they lag behind on our dataset.