STAR-Net: An Interpretable Model-Aided Network for Remote Sensing Image Denoising

2025-05-30

AI Summary

To address three key bottlenecks in remote sensing image denoising—limited interpretability, insufficient modeling of non-local self-similarity, and heavy reliance on manual regularization parameter tuning—this paper proposes STAR-Net and its robust variant, STAR-Net-S. Methodologically, we introduce the first ADMM-guided deep unrolling architecture enabling automatic learning of regularization parameters; integrate physics-based priors with data-driven learning to enhance model interpretability; and pioneer sparse tensor decomposition to explicitly capture non-local structural redundancy in remote sensing imagery, jointly leveraging low-rank tensor priors and robust non-Gaussian noise modeling. Extensive experiments on both synthetic and real-world remote sensing datasets demonstrate that STAR-Net and STAR-Net-S consistently outperform state-of-the-art methods, achieving significant gains in PSNR and SSIM. Moreover, the proposed models exhibit strong generalization capability and robustness across diverse noise distributions.

Technology Category

• Computer Vision: Remote Sensing / Geospatial AI
• Intelligent Robots: State Estimation
• Machine Learning: Deep Generative Models & Autoencoders

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: Graph neural networks and deep learning approaches for Web-related graphs
• Semantics and Knowledge: Methods, algorithms and applications for the development of semantic models, knowledge graphs and other forms of structured data models with machine-interpretable semantics

Abstract

Remote sensing image (RSI) denoising is an important topic in the field of remote sensing. Despite the impressive denoising performance of RSI denoising methods, most current deep learning-based approaches function as black boxes and lack integration with physical information models, leading to limited interpretability. Additionally, many methods may struggle with insufficient attention to non-local self-similarity in RSI and require tedious tuning of regularization parameters to achieve optimal performance, particularly in conventional iterative optimization approaches. In this paper, we first propose a novel RSI denoising method named sparse tensor-aided representation network (STAR-Net), which leverages a low-rank prior to effectively capture the non-local self-similarity within RSI. Furthermore, we extend STAR-Net to a sparse variant called STAR-Net-S to deal with the interference caused by non-Gaussian noise in original RSI for the purpose of improving robustness. Different from conventional iterative optimization, we develop an alternating direction method of multipliers (ADMM)-guided deep unrolling network, in which all regularization parameters can be automatically learned, thus inheriting the advantages of both model-based and deep learning-based approaches and successfully addressing the above-mentioned shortcomings. Comprehensive experiments on synthetic and real-world datasets demonstrate that STAR-Net and STAR-Net-S outperform state-of-the-art RSI denoising methods.

Problem

• Improves interpretability in RSI denoising models
• Enhances non-local self-similarity attention in RSI
• Automates regularization parameter tuning for optimal performance

Innovation

• Leverages low-rank prior for non-local similarity
• Uses ADMM-guided deep unrolling network
• Automatically learns regularization parameters