Towards Robust Learning based Image Denoising: An Overview

Image denoising is a fundamental yet challenging problem in the field of low-level vision tasks, as it directly affects the quality and usability of images in various applications. Over the years, numerous methods have been proposed to address this issue, with deep-learning based methods achieving p...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:2025 4th Asia Conference on Algorithms, Computing and Machine Learning (CACML) S. 1 - 5
1. Verfasser: Chen, Shiyue
Format: Tagungsbericht
Sprache:Englisch
Veröffentlicht: IEEE 28.03.2025
Schlagworte:
Adaptation models
Benchmark testing
Computational efficiency
Computational modeling
deep learning
Image denoising
Noise
Real-time systems
Robustness
Training
Usability
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:Image denoising is a fundamental yet challenging problem in the field of low-level vision tasks, as it directly affects the quality and usability of images in various applications. Over the years, numerous methods have been proposed to address this issue, with deep-learning based methods achieving particularly impressive performance. However, the robustness of these deep learning-based image denoising techniques is a crucial aspect that cannot be overlooked. Differences between testing and training conditions can significantly impact their performance in real-world scenarios, making it essential to develop methods that are both effective and adaptable. In recent years, many researchers have dedicated their efforts to improving the shortcomings of deep image denoising methods. Despite this, there has been a lack of systematic analysis and benchmarking of these methods, making it difficult to compare and evaluate their performance. To address this issue, we have conducted a comprehensive review of several deep learning-based image denoising methods which are all for natural images and classified them into four categories based on their different settings. For each category, we have selected representative methods and conducted benchmarking experiments using a public dataset, evaluating their performance both quantitatively and qualitatively. Through our experiments, we have observed some interesting trends and patterns in the performance of these methods. Based on our findings, we have provided detailed discussions and conclusions, highlighting the strengths and weaknesses of each method and offering insights into future research directions. Overall, this paper contributes to the advancement of image denoising techniques by providing a systematic review and benchmarking of existing methods.
DOI:10.1109/CACML64929.2025.11010929