USP-Mamba: Unmixing-Derived Spectral and Structural Prompting for Hyperspectral Image Super-Resolution

2026-08-03

AI Summary

This work addresses the challenge of hyperspectral image super-resolution, which requires enhancing spatial resolution while preserving spectral fidelity. Existing Mamba-based models suffer from degraded spatial adjacency due to one-dimensional scanning and lack explicit modeling of hyperspectral characteristics in their state space. To overcome these limitations, we propose USP-Mamba, the first framework to integrate spectral unmixing priors into the Mamba architecture. Our approach introduces a synergistic mechanism combining global spectral prompts with local structural prompts and employs a dual-path scanning strategy—Hilbert curve scanning for spatial continuity and semantic-guided neighborhood scanning for non-local semantic dependencies. Extensive experiments demonstrate that USP-Mamba significantly outperforms current state-of-the-art methods across multiple hyperspectral datasets, achieving superior reconstruction in both spatial detail and spectral fidelity.

Abstract

Hyperspectral image super-resolution aims to reconstruct high-resolution imagery while preserving dense spectral information. Recently, Mamba-based models have shown promising potential for this task by capturing long-range dependencies with linear computational complexity. Nevertheless, their causal sequence modeling requires two-dimensional hyperspectral features to be unfolded along predefined scanning orders, which disrupts spatial adjacency and restricts the effective propagation of contextual information. Moreover, state-space parameterization of existing models is predominantly derived from generic learned representations, without explicit alignment with the intrinsic characteristics of the hyperspectral image. To address this issue, we propose an Unmixing-derived Spectral and Structural Prompting Mamba framework, termed USP-Mamba, which adapts Mamba state evolution through composition-aware spectral priors and image-dependent structural prompts. Specifically, an unmixing-informed spectral prompt captures the global material composition of the input image and provides persistent conditioning throughout reconstruction. Injected into the Mamba sequence and progressively adapted across layers, it steers state evolution toward composition-consistent reconstruction. We introduce feature-level structural prompts comprising spatial and frequency components to provide image-dependent local guidance. The spatial prompt promotes structure-sensitive state encoding for local detail preservation, while the frequency prompt enables region-adaptive transitions between homogeneous regions and high-frequency details. Finally, complementary Hilbert and Semantic-Guided Neighboring scans preserve spatial continuity and strengthen non-local semantic dependency modeling. Extensive experiments on different datasets demonstrate that the proposed method consistently outperforms representative approaches.

Problem

• hyperspectral image super-resolution
• Mamba model
• spatial adjacency
• spectral information
• state-space modeling

Innovation

• spectral unmixing
• Mamba
• prompting mechanism
• hyperspectral super-resolution
• state-space modeling