Opendenoising: An Extensible Benchmark for Building Comparative Studies of Image Denoisers

Abstract

Image denoising has recently taken a leap forward due to machine learning. However, image denoisers, both expert-based and learning-based, are mostly tested on well-behaved generated noises (usually Gaussian) rather than on real-life noises, making performance comparisons difficult in real-world conditions. This is especially true for learning-based denoisers which performance depends on training data. Hence, choosing which method to use for a specific denoising problem is difficult.This paper proposes a comparative study of existing denoisers, as well as an extensible open tool that makes it possible to reproduce and extend the study. MWCNN is shown to outperform other methods when trained for a real-world image interception noise, and additionally is the second least compute hungry of the tested methods. To evaluate the robustness of conclusions, three test sets are compared. A Kendall's Tau correlation of only 60% is obtained on methods ranking between noise types, demonstrating the need for a benchmarking tool.

NoiseBreaker: Gradual Image Denoising Guided by Noise Analysis

“…It is composed of a combination of Poisson noise, Gaussian noise, and Bernoulli noise. In [20], authors compare methods designed for AWGN removal when trained and evaluated on more complex noises such as sequential mixtures of Gaussian and Bernoulli distributions. Experimental results show that denoising performances severely drop on complex noises even when using supervised learning methods like Denoising Convolutional Neural Network (DnCNN).…”

“…For each class, several state of the art denoising architectures have been compared to be used as primary denoisers. The comparison was done using the OpenDenoising benchmark tool [20]. The compared methods are Multi-level Wavelet Convolutionnal Neural Network (MWCNN) [6], Self-Guided Network (SGN) [15], Super-Resolution Residual Network (SRResNet) [13] and DnCNN [5].…”

Image denoising in the deep learning era

“…Driven by the availability of large datasets, rapid increases in computational power and, advances in algorithmic development for optimization of neural networks, deep learning has made impressive improvements in tackling many computer vision tasks [15][16][17][18]. In the context of image denoising, deep learning has attracted significant research interest and raised many new questions during the past recent years [2,[19][20][21] (Fig. 1).…”

“…2, different denoising paradigms have been proposed during the past years, however, most of these methods are tailored to specific contexts and are based on benchmark datasets that are not directly comparable. Additionally, some new benchmark datasets have been proposed which are not included in the existing reviews [2,[19][20][21]. This motivates us to examine the recent advances in this active area to deliver an overview as well as new perspectives for interesting research directions.…”