Spectrum-Driven Mixed-Frequency Network for Hyperspectral Salient Object Detection

Hyperspectral salient object detection (HSOD) aims to detect spectrally objects in hyperspectral images (HSIs). However, existing methods utilize spectral information by either converting HSIs into-color images or converging neural networks with clustering. We propose a approach that fully leverages the spectral characteristics by extracting distinct frequency components from the spectrum: low-frequency Spectral and high-frequency Spectral Edge. The Spectral Saliency approximates region of salient objects, while the Spectral Edge captures edge of salient objects. These two complementary components, crucial for, are derived by computing from the inter-layer spectral angular distance the Gaussian pyramid and the intra-neighborhood spectral angular gradients,. To effectively utilize this dual-frequency information, we a novel lightweight Spectrum-driven Mixed-frequency Network (SMN). incorporates two parameter-free plug-and-play operators, namely Spectral Generator and Spectral Edge Operator, to extract the Spectral Saliency Spectral Edge components from the input HSI independently. Subsequently, Mixed-frequency Attention module, comprised of two frequency-dependent, intelligently combines the embedded features of edge and saliency, resulting in a mixed-frequency feature representation., a saliency-edge-aware decoder progressively scales up the-frequency feature while preserving rich detail and saliency information accurate salient object prediction. Extensive experiments conducted on the-SOD benchmark and our custom dataset HSOD-BIT demonstrate that our SMN state-of-the-art methods regarding HSOD performance. Code and will be available at https://github.com/laprf/SMN.