Unsupervised Hyperspectral Denoising Based on Deep Image Prior and Least Favorable Distribution

Abstract

This paper considers the inverse problem under hyperspectral images (HSIs) denoising framework. Recently, it has been shown that deep learning is a promising approach to image denoising. However, deep learning to be effective usually needs a massive amount of training data. Moreover, in a practical scenario, HSIs may get contaminated by different kinds of noises such as Gaussian and/or sparse noise. Lately, it has been reported that the convolutional neural network (CNN), the core element used by deep image prior (DIP), is able to capture image statistical characteristics without the need of training, i.e., restore the clean image blindly. Nonetheless, there exists some performance gap between DIP and state-of-the-art methods in HSIs (e.g., low-rank models). By applying the Huber loss function (HLF), which is derived through a least favorable distribution in robust statistics, to DIP, we propose a novel unsupervised denoising algorithm, referred as to the HLF-DIP, free from pre-training and without involving any regularizer. Extensive experimental results are provided to demonstrate that the proposed HLF-DIP algorithm significantly outperforms seven state-of-the-art algorithms in both complexity (thanks to no regularization) and robustness against complex noise (e.g., mixed types of noises).

Polarimetric Imaging via Deep Learning: A Review

“…Of course, some models can achieve this function in an unsupervised way, such as the GAN network, but the performance of these methods is limited and always significantly worse than the performance of supervised methods. How to further enhance, especially by adding prior physical knowledge into training, the performance of unsupervised solutions is a burning problem [64,227].…”

“…To handle this issue, in 2022, Ref. [227] proposed a user-friendly unsupervised hyperspectral images denoising solution under a deep image prior framework. Extensive experimental results demonstrate that the method can preserve image edges and remove different noises, including mixed types (e.g., Gaussian noise, sparse noise).…”

“…Figure 29. Chronological evolution of the reviewed works.The references mentioned here are chronologically listed as 2014 [287], 2015 [188,189], 2016 [181,291], 2017 [52,184,278,304], 2018 [92,199,206,235,292,295], 2019 [70,91,97,140,212,221,225,234,236], 2020 [11,58,82,[173][174][175]222,273,303,305], 2021 [226,244,282,297], 2022 [22,64,211,227,231,258,259,294,306], and 2023 [210,…”

Beyond Supervised Learning in Remote Sensing: A Systematic Review of Deep Learning Approaches

“…In this study, Imamura et al, proposed a fully CNN network a trained it by manually adding noise the noisy HSI, while the model's objective was to generate the original image from the noisy one. In another study for HSI denoising, Faghih et al, (2022) proposed an unsupervised method using a deep image prior (DIP) UNet architecture [135]. They showed that DIP trained with Huber loss function can generate denoised HSI without needing training data.…”

Autonomous unsupervised image inpainting networks based on structural priors

“…Criminisi et al propose a texture synthesis method [11] ,which is mainly implemented by repeatedly computing the priority of the restored region [12] , propagating the texture information and updating the confidence level, but with the consequent repetition of steps, the confidence level decreases, leading to errors in the filling order.…”