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

Haijin Zeng,Wenfeng Kong,Xiaozhen Xie,Shuang Cui,Jifeng Ning

doi:10.1109/access.2020.2979809

Abstract

Hyperspectral images (HSIs) are usually corrupted by various noises during the image acquisition process, e.g., Gaussian noise, impulse noise, stripes, deadlines and many others. Such complex noise severely degrades the data quality, reduces the interpretation accuracy of HSIs, and restricts the subsequent HSI applications. In this paper, a spatial non-local and local rank-constrained low-rank regularized Plug-and-Play (NLRPnP) model is presented for mixed noise removal in HSIs. Specifically, we first divide HSIs into local overlapping patches. Local rank-constrained low-rank matrix recovery is adopted to effectively separate the low-rank clean HSI patches from the sparse noise and a part of Gaussian noise, and to significantly preserve local structure and detail information in HSIs. Then the spatial non-local based denoiser is introduced to promote the non-local self-similarity and obviously depress the Gaussian noise. Without increasing the difficulty of solving optimization problems, we combine the local and non-local based methods into the Plug-and-Play framework, and develop an efficient algorithm for solving the proposed NLRPnP model by using the alternating direction method of multipliers method. Finally, several experiments are conducted in both simulated and real data conditions to illustrate the better performance of the proposed NLRPnP model than the existing state-of-the-art denoising models.

Highlights

Hyperspectral images (HSIs) can provide spectral information about hundreds of continuous bands in the same scene, HSIs are widely used in many fields [1], [2]
EXPERIMENTAL RESULTS AND DISCUSSIONS the non-local and local rank-constrained low-rank regularized Plugand-Play (NLRPnP) method is applied to both simulation and real experiments for verifying its HSI denoising performance
In order to comprehensively evaluate the performance of our proposed model, three kinds of denoising methods are used for the comparison

Summary

INTRODUCTION

Hyperspectral images (HSIs) can provide spectral information about hundreds of continuous bands in the same scene, HSIs are widely used in many fields [1], [2]. For sufficiently using both the spatial and spectral prior information, many researchers attempt to integrate the VOLUME 8, 2020 different prior information by their according different mathematical formulations He et al [8] propose a LRTV model which uses the band-wised TV regularization to represent the spatial sparsity and simultaneously uses the nuclear norm minimization of HSIs’ Casorati matrices to describe the spectral low-rank prior. The regularization method is an effective and widely used method for solving such inverse problems It establishes the following regularized denoising framework by adding the prior information of unknown clear HSI and mixed noise, i.e., arg min J(X , S, N ) + β R(X ),. There are many state-of-the-art noise reducers that can be incorporated into this framework, such as BM3D [10], BM4D [31], NLM [32] and so on

LOCAL RPCA

OPTIMIZATION PROCEDURE

UPDATING THE LAGRANGIAN PARAMETERS

EXPERIMENTAL RESULTS AND DISCUSSIONS

CONCLUSION