USCTNet: A deep unfolding nuclear-norm optimization solver for physically consistent HSI reconstruction

2025-09-12

AI Summary

Reconstructing hyperspectral images (HSI) from a single RGB image is a severely ill-posed inverse problem, prone to physically inconsistent reconstructions due to unknown camera spectral sensitivity (CSS) and scene illumination. To address this, we propose USCTNet, a physics-driven deep unrolling network. First, we formulate a differentiable physical imaging model that jointly and explicitly estimates both CSS and scene illumination. Second, we design a data-adaptive low-rank subspace singular value thresholding (SVT) operator, circumventing full SVD computation to improve efficiency. Third, we introduce nuclear norm regularization in a learnable transform domain and realize end-to-end optimization via proximal gradient unrolling. Extensive experiments on multiple benchmark datasets demonstrate that USCTNet significantly outperforms existing RGB-to-HSI methods, achieving state-of-the-art performance in both reconstruction accuracy and spectral–chromatic physical consistency.

Technology Category

• Computer Vision: Low Level & Physics-based Vision
• Search and Optimization: Non-convex Optimization
• Constraint Satisfaction and Optimization: Distributed CSP/Optimization

Application Category

• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Economics, Online Markets and Human Computation: Data quality aspects of human-annotated datasets

Abstract

Reconstructing hyperspectral images (HSIs) from a single RGB image is ill-posed and can become physically inconsistent when the camera spectral sensitivity (CSS) and scene illumination are misspecified. We formulate RGB-to-HSI reconstruction as a physics-grounded inverse problem regularized by a nuclear norm in a learnable transform domain, and we explicitly estimate CSS and illumination to define the forward operator embedded in each iteration, ensuring colorimetric consistency. To avoid the cost and instability of full singular-value decompositions (SVDs) required by singular-value thresholding (SVT), we introduce a data-adaptive low-rank subspace SVT operator. Building on these components, we develop USCTNet, a deep unfolding solver tailored to HSI that couples a parameter estimation module with learnable proximal updates. Extensive experiments on standard benchmarks show consistent improvements over state-of-the-art RGB-based methods in reconstruction accuracy. Code: https://github.com/psykheXX/USCTNet-Code-Implementation.git

Problem

• Reconstructing hyperspectral images from single RGB inputs
• Addressing physical inconsistency from misspecified camera and illumination
• Avoiding computational cost of full singular value decompositions

Innovation

• Deep unfolding nuclear-norm optimization solver
• Data-adaptive low-rank subspace SVT operator
• Physics-grounded inverse problem with learnable transforms