[論文レビュー] Unleashing the Power of Self-Supervised Image Denoising: A Comprehensive Review

Dan Zhang, Fangfang Zhou+5 more|arXiv (Cornell University)|Aug 1, 2023

Image and Signal Denoising Methods被引用数 8

ひとこと要約

A comprehensive survey of self-supervised image denoising methods, categorized into General, Blind Spot Network (BSN)-based, and Transformer-based approaches, with theoretical insights and experimental analysis.

ABSTRACT

The advent of deep learning has brought a revolutionary transformation to image denoising techniques. However, the persistent challenge of acquiring noise-clean pairs for supervised methods in real-world scenarios remains formidable, necessitating the exploration of more practical self-supervised image denoising. This paper focuses on self-supervised image denoising methods that offer effective solutions to address this challenge. Our comprehensive review thoroughly analyzes the latest advancements in self-supervised image denoising approaches, categorizing them into three distinct classes: General methods, Blind Spot Network (BSN)-based methods, and Transformer-based methods. For each class, we provide a concise theoretical analysis along with their practical applications. To assess the effectiveness of these methods, we present both quantitative and qualitative experimental results on various datasets, utilizing classical algorithms as benchmarks. Additionally, we critically discuss the current limitations of these methods and propose promising directions for future research. By offering a detailed overview of recent developments in self-supervised image denoising, this review serves as an invaluable resource for researchers and practitioners in the field, facilitating a deeper understanding of this emerging domain and inspiring further advancements.

研究の動機と目的

• Motivate and map the challenges of denoising without clean ground-truth data in real-world settings.
• Categorize self-supervised denoising methods into three classes and summarize their principles and applications.
• Provide theoretical analyses and practical evaluations across datasets to guide researchers and practitioners.

提案手法

• Classify self-supervised denoising methods into General, BSN-based, and Transformer-based categories.
• Summarize representative methods and their core principles, including masking strategies and score-based approaches.
• Discuss datasets, evaluation metrics, and both qualitative and quantitative results to compare methods.

実験結果

リサーチクエスチョン

• RQ1 What are the principal self-supervised denoising paradigms and their underlying assumptions?
• RQ2 How do General, BSN-based, and Transformer-based methods compare in terms of theoretical foundations and practical performance?
• RQ3 What datasets and metrics best reveal the strengths and limitations of self-supervised denoising methods?
• RQ4 What are the key limitations and promising directions for future research in self-supervised image denoising.

主な発見

• The paper presents a three-way taxonomy of self-supervised denoising methods: General, BSN-based, and Transformer-based.
• It provides concise theoretical analyses and describes practical applications for each category.
• Experimental results are discussed to assess effectiveness across various datasets and denoising tasks.
• The review highlights current limitations and suggests directions for future research in self-supervised denoising.