ZERO-SHOT HYPERSPECTRAL IMAGE DENOISING WITH SELF-COMPLETION WITH PATTERNED MASKS

Hyperspectral images (HSI) are data with higher spectral resolution than RGB images and are used for various tasks. HSI images are prone to degradation due to noise during imaging, and the high cost of imaging makes it very time-consuming to prepare large amounts of data. Therefore, it is often difficult to learn HSI using general deep learning-based methods that use degraded and non-degraded image pairs. Many methods that do not require training images have been proposed. With the recent development of self-supervised learning in image restoration tasks, learning-based image restoration methods that do not require non-degraded images for RGB images have been successful. This paper proposes a zero-shot HSI deep denoising method based on self-supervised image restoration. The feature of this method is that it learns only from a single observed HSI. Numerical experiments show that this method has higher restoration accuracy than conventional zero-shot methods.