Frequency-Enhanced Spatial-Spectral Network for Hyperspectral and Multispectral Image Fusion

Fusion of hyperspectral images (HSIs) and multispectral images (MSIs) is an effective approach to enhance the spatial resolution of HSIs. Existing deep learning-based methods have achieved excellent results, but they often overlook the important role of frequency information. To address this issue, we propose a frequency-enhanced spatial-spectral fusion network for HSIs and MSIs fusion, called FSSNet. First, we designed a frequency information extraction module (FIEM) to extract frequency features. Subsequently, the extracted frequency features are integrated into spatial-spectral features through the stripe feature enhancement module (SFEM). Finally, a deep residual reconstruction module (DRRM) is introduced to integrate all features and reconstruct the high-resolution hyperspectral image (HR-HSI). Qualitative and quantitative experiments on the Chikusei dataset demonstrate the effectiveness of the proposed method compared with five state-of-the-art (SOTA) methods.