WaveHSR: A Wavelet-Based Microscopic Hyperspectral Image Super-Resolution Network

Microscopic hyperspectral images (MHSIs) provide rich spectral information but often suffer from low spatial resolution due to hardware limitations. To address this, we propose WaveHSR, a novel wavelet-based microscopic hyperspectral image super-resolution network. The proposed method integrates both hyperspectral and multispectral images (MSIs), leveraging wavelet transforms to decompose them into low-frequency and high-frequency components. Through cross-modality fusion blocks, we efficiently combine the spectral richness of MHSIs with the spatial details of MSIs in the frequency domain, capturing complementary information from both modalities. Moreover, our network employs parallel sub-networks to independently process different spectral bands, ensuring robust detail preservation. Extensive experiments on a public dataset demonstrate that WaveHSR outperforms state-of-the-art methods in both quantitative metrics and visual quality.