Abstract:Hyperspectral image (HSI) denoising is an essential preprocessing step for downstream applications. Fully characterizing the spatial-spectral priors of HSI is crucial for HSI denoising tasks. In recent years, denoising methods based on low-rank subspaces have garnered attention. Within the low-rank decomposition framework (LRDF), the restoration of HSIs can be formulated as a problem of restoring two subspace factors. Since the rank of the HSI data has been predetermined by LRDF, subspace-based methods have already characterized the spectral low-rankness information. Next, subspace-based methods only need to encode spatial priors for HSIs. Existing subspace-based methods either rely on a manual-designed regularization or a pre-trained deep neural network. The former fails to fully capture the intrinsic priors of the HSI, while the latter may encounter generalization issues. Inspired by the unsupervised deep image prior (DIP) technique, this article proposes a double subspace deep prior (DSDP) model to track the mentioned issues. In this model, the two subspace factors are parallelly represented by two deep neural networks. By incorporating popular attention modules into classical convolutional neural networks, the well-designed subspace factor neural network can effectively capture the deep prior of the two subspace factors separately from each HSI in an unsupervised manner. Additionally, the total variation (TV) regularizer is added to constrain the generation of the subspace factor neural network, and further to ensure the effectiveness and robustness of the parameter learning process. Extensive experiments demonstrate that our method outperforms a series of competing methods.

Denoising of Hyperspectral Image Using Low-Rank Matrix Factorization.

Hyperspectral Imagery Denoising Via Reweighed Sparse Low-Rank Nonnegative Tensor Factorization

Denoising of Hyperspectral Images Using Nonconvex Low Rank Matrix Approximation.

Group Sparse Nonnegative Matrix Factorization for Hyperspectral Image Denoising

Denoising of Hyperspectral Images Using Group Low-Rank Representation.

Hyperspectral Image Denoising Via Spectral and Spatial Low-Rank Approximation

Hyperspectral Image Denoising Via Sparse Representation and Low-Rank Constraint

Hyperspectral Image Denoising with Partially Orthogonal Matrix Vector Tensor Factorization

Multitask Sparse Nonnegative Matrix Factorization for Joint Spectral–Spatial Hyperspectral Imagery Denoising

Hyperspectral Image Denoising Via Double Subspace Deep Prior

Hyperspectral Restoration Via $L_0$ Gradient Regularized Low-Rank Tensor Factorization

Hyperspectral Image Denoising Via Sparsity and Low Rank

Hyperspectral Image Denoising via Low-Rank Representation and CNN Denoiser

Low-rank Bayesian Tensor Factorization for Hyperspectral Image Denoising.

Hyperspectral Image Denoising Via Combined Non-Local Self-Similarity and Local Low-Rank Regularization

Hyperspectral Images Denoising via Nonconvex Regularized Low-Rank and Sparse Matrix Decomposition

Learning a Model-Based Deep Hyperspectral Denoiser from a Single Noisy Hyperspectral Image.

A hyperspectral image denoising method based on land cover spectral autocorrelation

Hyperspectral Image Denoising Based on Nonlocal Low-Rank and TV Regularization

Research on Denoising Methods for Hyperspectral Images Based on Low-Rank Theory and Sparse Representation

Hyperspectral Image Denoising Using Nonconvex Fraction Function.