A novel non-convex low-rank tensor approximation model for hyperspectral image restoration

作者：

Highlights：

• A novel low-Tucker-rank regularization ϵ-norm for capturing the structure of the low-rank tensor is constructed.

• An ϵ-norm based non-convex low-rank tensor approximation model for mixed noise removal is proposed.

• An efficient algorithm based on the augmented Lagrangian multipliers method for solving the proposed non-convex optimization problem is developed.

• Extensive experiments on simulated and real experiments are conducted.

摘要

•A novel low-Tucker-rank regularization ϵ-norm for capturing the structure of the low-rank tensor is constructed.•An ϵ-norm based non-convex low-rank tensor approximation model for mixed noise removal is proposed.•An efficient algorithm based on the augmented Lagrangian multipliers method for solving the proposed non-convex optimization problem is developed.•Extensive experiments on simulated and real experiments are conducted.

论文关键词：Low-Rank tensor approximation,Mixed noise,Non-convex optimization,Restoration,Hyperspectral image

论文评审过程：Received 4 November 2019, Revised 1 April 2021, Accepted 2 May 2021, Available online 24 May 2021, Version of Record 24 May 2021.

论文官网地址：https://doi.org/10.1016/j.amc.2021.126342