UNSURE: self-supervised learning with Unknown Noise level and Stein's Unbiased Risk Estimate

2024-09-03

AI Summary

This work addresses image reconstruction under unknown noise levels and without ground-truth labels. To this end, we propose a novel self-supervised denoising framework. Methodologically, we introduce a SURE-driven self-supervised loss, a noise-insensitive parameterized noise modeling scheme, and a differentiable inverse problem solver. Our key contribution is the first extension of Stein’s Unbiased Risk Estimator (SURE) theory to scenarios with unknown noise priors—establishing a theoretically grounded trade-off between model expressivity and robustness, thereby overcoming the classical SURE requirement of precise noise characterization. Extensive experiments on CT reconstruction, image denoising, and super-resolution demonstrate that our approach significantly outperforms state-of-the-art self-supervised methods—including Noise2Self and Noise2Void—and approaches the performance upper bound of fully supervised learning.

Technology Category

• Machine Learning: Unsupervised & Self-Supervised Learning
• Computer Vision: Adversarial Attacks & Robustness
• Reasoning under Uncertainty: Stochastic Optimization

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs
• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation

Abstract

Recently, many self-supervised learning methods for image reconstruction have been proposed that can learn from noisy data alone, bypassing the need for ground-truth references. Most existing methods cluster around two classes: i) Stein's Unbiased Risk Estimate (SURE) and similar approaches that assume full knowledge of the distribution, and ii) Noise2Self and similar cross-validation methods that require very mild knowledge about the noise distribution. The first class of methods tends to be impractical, as the noise level is often unknown in real-world applications, and the second class is often suboptimal compared to supervised learning. In this paper, we provide a theoretical framework that characterizes this expressivity-robustness trade-off and propose a new approach based on SURE, but unlike the standard SURE, does not require knowledge about the noise level. Throughout a series of experiments, we show that the proposed estimator outperforms other existing self-supervised methods on various imaging inverse problems.

Problem

• Image Denoising
• Unknown Noise Level
• Blind Denoising

Innovation

• SURE Principle
• Noise-level Agnostic
• Image Deblurring