基于低秩稀疏分解的湍流退化图像序列的盲去卷积算法  

A multi-frame blind deconvolution algorithm based on low-rank decomposition for turbulence-degraded images

作　　者：张姣 李俊山[1] 李成 胡双演[1] 汪晓建 

机构地区：[1]第二炮兵工程大学,陕西西安710025 [2]武警工程大学,陕西西安710086

出　　处：《光电子．激光》2015年第7期1374-1380,共7页Journal of Optoelectronics·Laser

基　　金：国家自然科学基金(61175120)资助项目

摘　　要：针对湍流退化图像序列存在像偏移、像抖动和像模糊的问题,提出一种基于低秩稀疏分解和多帧去卷积的图像复原算法。首先分析大气湍流下图像序列的退化特征,然后在低秩稀疏分解的思想下,采用非增广拉格朗日乘子(IALM)法优化由低秩矩阵的核范数和稀疏矩阵的Frobenius范数之和构成的目标函数,将湍流退化序列分解为低秩稳像和稀疏湍流两部分;最后利用多帧去卷积算法复原对齐的稳像。实验结果表明,本文算法能够有效校正湍流像素偏移,在提高复原质量和速度方面取得了明显的效果。In order to correct pixel distortion and reduce time-varying blur,a novel blind restoration algorithm based on low-rank decomposition is proposed in this paper,which is capable of restoring a highquality image from the turbulence-degraded image sequence.The paper firstly analyzes the infrared image characteristics at the atmospheric turbulence.Next,it decomposes the turbulence sequence into two components of the stabilized background,the turbulence and noise through low-rank decomposition.The decomposition is simplified into a minimization of nuclear norm and Frobenius norm,optimized by inexact augmented Lagrange multiplier（IALM）.Finally,a multi-frame blind deconvolution algorithm is implemented to deblur the stabilized images,generating the final output.Experimental results verify the effectiveness of the algorithm.It can effectively alleviate blur and distortions,improve visual quality and recovery speed significantly.

关 键 词：图像高复原 低秩分解 非增广拉格朗日乘子法(IALM) 大气湍流退化 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]