基于低秩先验非局部匹配的高光谱图像恢复  

Hyper-spectral image restoration based on low rank a priori nonlocal matching

作　　者：刘春茂[1] LIU Chun-mao(School of Electronic Information Engineering,Henan Polytechnic College,Nanyang 473000,China)

机构地区：[1]河南工业职业技术学院电子信息工程学院,河南南阳473000

出　　处：《激光与红外》2023年第2期302-312,共11页Laser & Infrared

基　　金：全国高等院校计算机基础教育研究会2021年度规划项目(No.2021-AFCEC-208)资助。

摘　　要：为了实现未知退化类型条件下的图像恢复,提出了一种基于低秩先验非局部匹配的高光谱图像恢复方法。该恢复策略将结构相似性指数度量作为数据保真度项,核范数作为正则化项,从而能够使用单一算法解决延迟和信号相关的退化形式,能够处理加性高斯噪声、与信号相关的泊松噪声、混合泊松-高斯噪声,并且能够恢复被截止期和条纹损坏的高光谱图像。通过多个数据集实验结果证明了提出方法具有较好的恢复效果。In order to realize image restoration under unknown degradation type conditions, a hyper-spectral image restoration method based on low-rank a priori non-local matching is proposed.Using the structural similarity index as the data fidelity term and the kernel norm as the regularization term, a single algorithm can be used to solve the degradation forms of delay and signal correlation, deal with additive Gaussian noise, signal related Poisson noise and mixed Poisson Gaussian noise, and recover hyper-spectral images damaged by deadline and fringe.The experimental results of several data sets show that the proposed method has good recovery effect.

关 键 词：图像恢复 高光谱图像 结构相似性指数度量 泊松噪声 

分 类 号：TP391.14[自动化与计算机技术—计算机应用技术]