Abstract:In this paper, a novel hyperspectral denoising method is proposed, aiming at restoring clean images from images disturbed by complex noise. Previous denoising methods have mostly focused on exploring the spatial and spectral correlations of hyperspectral data. The performances of these methods are often limited by the effective information of the neighboring bands of the image patches in the spectral dimension, as the neighboring bands often suffer from similar noise interference. On the contrary, this study designed a cross-band non-local attention module with the aim of finding the optimal similar band for the input band. To avoid being limited to neighboring bands, this study also set up a memory library that can remember the detailed information of each input band during denoising training, fully learning the spectral information of the data. In addition, we use dense connected module to extract multi-scale spatial information from images separately. The proposed network is validated on both synthetic and real data. Compared with other recent hyperspectral denoising methods, the proposed method not only demonstrates good performance but also achieves better generalization.

What problem does this paper attempt to address?

### Problems the Paper Aims to Solve This paper aims to address the problem of denoising hyperspectral images (HSIs). Specifically, the paper proposes a new hyperspectral denoising method with the goal of recovering clean images from those affected by complex noise. #### Background and Challenges Traditional denoising methods mainly focus on exploring the spatial and spectral correlations of hyperspectral data. However, these methods are often limited by the effective information in adjacent bands in the spectral dimension, as adjacent bands are usually affected by similar noise. Therefore, existing methods have limited performance when dealing with complex noise. #### Features of the New Method To overcome the above issues, the paper designs a Cross-Band Non-Local Attention Module to find the best similar bands of the input band. Additionally, the study establishes a Memory Library that can remember the detailed information of each input band during the denoising training process, thereby fully learning the spectral information of the data. Moreover, the method utilizes a Dense Connected Module to extract multi-scale spatial information from the image. #### Method Overview 1. **First Stage**: Use the Dense Connected Module (DCM) to extract multi-scale spatial features of the current input band and its adjacent bands. 2. **Second Stage**: Includes two modules: - Non-Local Spectral Attention Module, used to extract useful information from adjacent band features, supplement the current band information, and remove noise. - Global Spectral Memory-Augmented Module, which directly queries all useful information from the global memory library to restore the current band. By combining the outputs of these two modules and further processing through convolutional layers, and finally integrating the original features, the model can capture different spectral aspects of the input data and effectively integrate them. Finally, an upsampling module is used to fuse information from three scales and output the denoised result. ### Experimental Results The paper conducts experimental validation on both synthetic and real data. The results show that this method performs excellently in various noise environments, outperforming several other mainstream hyperspectral denoising methods. Particularly in complex noise situations, this method shows more stable performance. The visualization results also demonstrate that this method has better reconstruction effects in detail and edge regions.

Memory Augmentation and Non-Local Spectral Attention for Hyperspectral Denoising

Hyperspectral Remote Image Denoising Based on Multitask Nonnegative Dictionary Learning

A Hyperspectral Image Classification Method Based on the Nonlocal Attention Mechanism of a Multiscale Convolutional Neural Network.

Attention in Attention for Hyperspectral with High Spatial Resolution (H) Image Classification

Attention-Based Octave Network for Hyperspectral Image Denoising

Nonlocal Spatial–Spectral Neural Network for Hyperspectral Image Denoising

Multitask Sparse Representation Model-Inspired Network for Hyperspectral Image Denoising

MAC-Net: Model-Aided Nonlocal Neural Network for Hyperspectral Image Denoising.

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

Spatial-spectral convolutional sparse neural network for hyperspectral image denoising

Self-Supervised Learning Hyperspectral Image Denoiser with Separated Spectral-Spatial Feature Extraction.

A hyperspectral image denoising method based on land cover spectral autocorrelation

Spectral-Spatial Fused Attention Network for Hyperspectral Image Classification

Hybrid Spatial-spectral Neural Network for Hyperspectral Image Denoising

ADRN: Attention-based Deep Residual Network for Hyperspectral Image Denoising

Hyperspectral Image Denoising with Composite Regularization Models.

Hybrid Convolutional and Attention Network for Hyperspectral Image Denoising

Dynamic Adaptive Attention Guided Self-Supervised Single Remote Sensing Image Denoising

Hyperspectral Image Classification Based on a 3D Octave Convolution and 3D Multiscale Spatial Attention Network

Spatial-Frequency Attention for Image Denoising

Three-Dimension Spatial–Spectral Attention Transformer for Hyperspectral Image Denoising