MDHR-Net: Multimodal Hyperspectral Image Dynamic Fusion Super-Resolution

Hyperspectral image super-resolution fusion (HSRF) technique generates high-resolution hyperspectral images (HRHSI) by fusing low-resolution hyperspectral images (LR-HSI) with high-resolution multispectral images (HR-MSI). However, the traditional fusion method often affects the subsequent quantitative analysis due to insufficient spectral fidelity when improving spatial resolution. In this paper, a multimodal dynamic hybrid optimized network architecture is proposed to capture high-frequency and low-frequency information through adaptive feature enhancement blocks, thus significantly improving the quality of the fused images, effectively balancing spectral fidelity and spatial resolution, separating spectral and spatial features by using spatially decoupled hybrid attention blocks, and optimizing the resolution through the dynamic fusion mechanism.