S2CycleDiff: Spatial-Spectral-Bilateral Cycle-Diffusion Framework for Hyperspectral Image Super-resolution

• DOI: 10.1609/aaai.v38i5.28262
• Publication Date: 2024-03-25T09:40:10Z
• Authors: Jiahui Qu; Jie He; Wenqian Dong; Jingyu Zhao

Hyperspectral image super-resolution (HISR) is a technique that can break through the limitation of imaging mechanism to obtain hyperspectral (HSI) with high spatial resolution. Although some progress has been achieved by existing methods, most them directly learn spatial-spectral joint mapping between observed images and target high-resolution HSI (HrHSI), failing fully reserve spectral distribution low-resolution (LrHSI) multispectral imagery (HrMSI). To this end, we propose spatial-spectral-bilateral cycle-diffusion framework (S2CycleDiff) for HISR, which step-wise generate HrHSI fidelity learning conditional processes bilaterally. Specifically, customized designed as backbone achieve repeated refinement, wherein spatial/spectral guided pyramid denoising (SGPD) module seperately takes HrMSI LrHSI guiding factors details injection correction. The outputs are fed into complementary fusion block integrate desired HrHSI. Experiments have conducted on three widely used datasets demonstrate superiority proposed method over state-of-the-art HISR methods. code available at https://github.com/Jiahuiqu/S2CycleDiff.