Spectral-Spatial Synergistic Guided Network for Hyperspectral Salient Object Detection

2026-07-23

AI Summary

This work addresses the vulnerability of existing hyperspectral salient object detection methods to external factors such as illumination variations, which often misinterpret extrinsic spectral changes as intrinsic target characteristics, resulting in fragile representations and noisy predictions. To overcome these limitations, we propose S3GNet, a lightweight and efficient network that leverages a parameter-free spectral structure-aware module to extract robust intrinsic features. The architecture incorporates a flow-aware attention mechanism to enable cross-stream collaboration between spectral and spatial information and introduces a progressive gated refinement decoder to fuse multi-scale features for enhanced boundary accuracy. The proposed method achieves state-of-the-art performance while maintaining computational efficiency, demonstrating significant improvements in both robustness and detection precision.

Abstract

Hyperspectral salient object detection aims to identify visually salient regions from hyperspectral images. Existing methods often fail because they fundamentally misunderstand the data, confusing incidental spectral variations caused by external factors such as illumination with essential spectral differences caused by the intrinsic material properties of the object. This leads to fragile representations and noisy predictions. To this end, we propose a lightweight and efficient Spectral-Spatial Synergistic Guided Network (S3GNet), with structure perception as the core, to build a closed-loop information flow around spectrum robust modeling, cross-stream co-perception and multi-scale refinement decoding. S3GNet introduces a parameter-free Spectral Structure-Aware Module that leverages spectral derivatives and regional hierarchical modeling to extract intrinsic features of robustness against illumination variations. Our Stream-Aware Attention Module achieves effective spectral-spatial collaboration through inter-stream global interaction and intra-stream spatial guidance. Furthermore, a Progressive Gated Refinement Decoder ensures precise object boundaries and detail recovery by optimally integrating multi-scale features. Experimental results show that S3GNet achieves superior performance in both computational efficiency and detection accuracy compared to existing methods.

Problem

• Hyperspectral Salient Object Detection
• Spectral Variations
• Illumination Robustness
• Intrinsic Material Properties
• Noisy Predictions

Innovation

• Spectral-Spatial Synergy
• Hyperspectral Salient Object Detection
• Illumination-Robust Modeling
• Stream-Aware Attention
• Progressive Gated Refinement