Abstract
A new nonconvex smooth rank approximation model is proposed to deal with HSI mixed noise in this paper. The low-rank matrix with Laplace function regularization is used to approximate the nuclear norm, and its performance is superior to the nuclear norm regularization. A new phase congruency lp norm model is proposed to constrain the spatial structure information of hyperspectral images, to solve the phenomenon of “artificial artifact” in the process of hyperspectral image denoising. This model not only makes use of the low-rank characteristic of the hyperspectral image accurately, but also combines the structural information of all bands and the local information of the neighborhood, and then based on the Alternating Direction Method of Multipliers (ADMM), an optimization method for solving the model is proposed. The results of simulation and real data experiments show that the proposed method is more effective than the competcing state-of-the-art denoising methods.
Highlights
Due to the influence of many factors in the process and transmission of hyperspectral images, the acquired hyperspectral images often contain some complex mixed noises, including gauss noise, salt-and-pepper noise, and dead-line noise
We introduce the proposed proxy into a low-rank tensor separation model and solve the model by using the Alternating Direction Method of Multipliers (ADMM) algorithm, which can effectively complete the missing elements in the tensor, to achieve the goal of hyperspectral image denoising
To better illustrate the superiority of the combination of the nonconvex smooth rank approximation model and the lp norm constraint, the validity of the proposed method is verified by simulation and real experiments, and the quantitative and visual performance of the four advanced HSI denoising methods are compared with the denoising results of this method. ese methods include blockmatched 4D filtering (BM4D), Nonlrma [14], global spatial something spectral total variation (GSSTV) [15], and weighted group sparse regularized low-rank tensor decomposition (WGLRTD). e code of all comparison methods is Matlab Code
Summary
Due to the influence of many factors in the process and transmission of hyperspectral images, the acquired hyperspectral images often contain some complex mixed noises, including gauss noise, salt-and-pepper noise, and dead-line noise. HSI data in the real-world are often mixed into the real data by various noise combinations To solve this problem, some multidimensional methods are proposed to deal with both spectral and spatial information. Different from the traditional method, LRMR can process different types of noise without any prior information of noise; many methods of HSI image denoising [4–7] based on low rank and spectral correlation have been proposed. The low-rank model based on nuclear norm approximation is not a good approximation rank function but lacks image edge sparsity and structure smoothness regularization. Erefore, in this paper, a new phase congruency lp norm constraint is proposed to constrain the spatial structure information of hyperspectral images, to solve the phenomenon of “artificial artifact” in the process of hyperspectral image denoising Aiming at the shortcomings of the existing subband TV regularization and low-rank tensor denoising methods, which cannot effectively utilize the local neighborhood information, it is easy to cause the “artificial artifact” phenomenon, especially in the curved edge. e gradient cannot accurately describe the true structure of the image. erefore, in this paper, a new phase congruency lp norm constraint is proposed to constrain the spatial structure information of hyperspectral images, to solve the phenomenon of “artificial artifact” in the process of hyperspectral image denoising
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
