An adaptive deep denoising approach for Chandrayaan-2 IIRS Data with SURE Loss

This paper addresses the denoising of hyperspectral images acquired from the Chandrayaan-2 Imaging Infrared Spectrometer (IIRS) dataset, which are susceptible to various sources of noise, including sensor noise and other distortions. The presence of these noises presents challenges during data interpretation and analysis. To tackle this issue, we propose a deep learning-based approach that integrates a Convolutional Neural Network (CNN) architecture with the Stein's unbiased risk estimate (SURE) loss function. In our study, we extend this approach to a supervised framework, offering two distinct variants: one based on the traditional unsupervised approach and another incorporating a weighted loss function. By training our models using the CNN-SURE framework within the supervised context, we aim to effectively mitigate noise and enhance the quality of hyperspectral images. We demonstrate the efficacy of our approach in reducing noise artifacts and improving data interpretability. The denoised hyperspectral images obtained through our method exhibit significant potential for a wide range of applications in remote sensing and related fields