Unsupervised hyperspectral pansharpening via low-rank diffusion model

Xiangyu Rui, Xiangyong Cao, Li Pang, Zeyu Zhu, Zongsheng Yue, Deyu Meng

Abstract: Highlights•An unsupervised method named PLRDiff is proposed for hyperspectral pansharpening.•A pre-trained unconditional diffusion model is used to characterize image prior.•An easy-to-use low-rank factorization strategy is proposed for efficient sampling.•PLRDiff can finely generalize to different datasets and degradation conditions.