Hyperspectral Image Joint Super-resolution via Implicit Neural Representation

Hyperspectral image (HSI) joint super-resolution (SR) in both spatial and spectral dimensions is an area of increasing interest in HSI processing. Although recent advances in deep learning (DL) frameworks have greatly improved the performance of joint SR reconstruction, existing methods learn discrete representations of HSI, ignoring real-world signals' continuous nature. In this paper, we propose a joint SR method based on implicit neural representation (INR), which learns local continuous representations of high spatial resolution hyperspectral images from the discrete inputs. Experiments on joint SR demonstrate that our method can achieve superior performance in comparison with state-of-the-art methods.