Spectral Dynamic Attention Network for Hyperspectral Image Super-Resolution

2026-04-29

AI Summary

This work addresses the challenges in hyperspectral image super-resolution, particularly severe spectral redundancy and the limited nonlinear modeling capacity of conventional feedforward networks. To overcome these issues, the authors propose SDANet, a novel framework that introduces a Dynamic Channel-Sparse Attention (DCSA) module to enable data-dependent sparse spectral interactions. Additionally, a Frequency-Enhanced Feedforward Network (FE-FFN) is designed to jointly model spatial and frequency-domain information. The proposed method achieves state-of-the-art performance on two benchmark datasets while maintaining high computational efficiency.

Abstract

Hyperspectral image super-resolution is essential for enhancing the spatial fidelity of HSI data, yet existing deep learning methods often struggle with substantial spectral redundancy and the limited non-linear modeling capacity of standard feed-forward networks (FFNs). To address these challenges, we propose Spectral Dynamic Attention Network (SDANet), a framework designed to adaptively suppress redundant spectral interactions. SDANet integrates two key components: 1) Dynamic Channel Sparse Attention (DCSA) module that computes channel-wise correlations and selectively preserves the most informative attention responses through dynamic and data-dependent sparsification. 2) Frequency-Enhanced Feed-Forward Network (FE-FFN) that jointly models spatial and frequency-domain representations to enhance non-linear expressiveness. Extensive experiments on two benchmark datasets demonstrate that SDANet achieves state-of-the-art HISR performance while maintaining competitive efficiency. The code will be made publicly available at https://github.com/oucailab/SDANet.

Problem

• hyperspectral image super-resolution
• spectral redundancy
• non-linear modeling
• deep learning

Innovation

• Dynamic Channel Sparse Attention
• Frequency-Enhanced Feed-Forward Network
• Hyperspectral Image Super-Resolution
• Spectral Redundancy Suppression
• Spectral Dynamic Attention