基于加速正则化RL算法的大气湍流退化图像盲复原方法

大气与环境光学学报 ›› 2011, Vol. 6 ›› Issue (5): 342-350.

基于加速正则化RL算法的大气湍流退化图像盲复原方法

李勇，范承玉，时东锋

（中国科学院安徽光学精密机械研究所中国科学院大气成分与光学重点实验室, 安徽合肥 230031）

收稿日期:2011-04-18 修回日期:2011-04-27 发布日期:2011-09-09
通讯作者: 范承玉（1965---），男，研究员，博士生导师，主要从事激光大气传输强湍流效应及其自适应光学校正方面的研究。Email：cyfan@aiofm.ac.cn E-mail:liyong2068@hotmail.com
作者简介:李勇(1985- ), 男, 湖南衡阳人, 硕士, 从事大气湍流退化图像处理方面的研究：liyong2068@hotmail.com
基金资助:
国家高技术研究发展计划（A825021,A825011）、和中科院合肥物质科学研究院计算中心项目（0330405002-7）资助

Atmospheric Turbulence-Degraded Image Blind Restoration Method Using the Accelerated and Regularized RL Algorithm

LI Yong, FAN Cheng-yu, SHI Dong-feng

( Key Laboratory of Atmospheric Composition and Optical Radiation, Anhui Institute of Optics and Fine Mechanics, Chinese Academy of Sciences, Hefei, 230031, China)

Received:2011-04-18 Revised:2011-04-27 Online:2011-09-09

摘要/Abstract

摘要：

提出了一种基于加速正则化Richardson-Lucy（RL）算法的大气湍流退化图像盲复原方法（AccRLTV-IBD）。在总变分（TV）正则化RL算法的基础上，引入二阶矢量外推加速技术对其进行加速，形成加速正则化RL（AccRLTV）算法，并将该算法应用到迭代盲目反卷积（IBD）算法中。使用长曝光大气湍流光学传递函数（OTF）的物理模型或根据图像来获取初始的点扩散函数（PSF），在灰度平均梯度（gray Mean Grads, GMG）的基础上定义了一个相对灰度平均梯度（relative Gray Mean Grads, RGMG）参数作为无参考图像复原质量的评价标准。模拟图像和实际湍流退化图像复原结果表明，基于RL的IBD算法要优于基于Wiener滤波的IBD算法，并且与RL-IBD算法相比，AccRLTV-IBD收敛速度更快，复原效果更好。

关键词: 图像复原, 大气湍流退化图像, 迭代盲目反卷积, RL算法, 矢量外推加速技术, 总变分正则化

Abstract:

A turbulence-degraded image blind restoration method (AccRLTV-IBD) using accelerated and regularized Richardson-Lucy(RL) algorithm is proposed. The accelerated and regularized RL (AccRLTV) algorithm is based on total variation (TV) regularized RL algorithm and vector extrapolation acceleration technique. The observed image or the physical model of long exposure atmospheric turbulence optical transfer function (OTF) is used to obtain the initial point spread function (PSF), and a non-reference restored image quality evaluation criterion called relative gray mean grads (RGMG) is defined on the basis of other evaluation criterion called gray mean grads (GMG). The restored images of simulation data and real turbulence-degraded data show that the IBD method based on RL is better than the IBD method based on Wiener, and AccRLTV-IBD, compared with RL-IBD algorithm, has greater convergence rate and better restoration result.

Key words: image restoration, atmospheric turbulence-degraded image, iterative blind deconvolution, Richardson-Lucy algorithm, vector extrapolation acceleration technique, total variation regularization

中图分类号:

TP391.41

李勇范承玉时东锋. 基于加速正则化RL算法的大气湍流退化图像盲复原方法[J]. 大气与环境光学学报, 2011, 6(5): 342-350.

LI Yong, FAN Cheng-yu, SHI Dong-feng. Atmospheric Turbulence-Degraded Image Blind Restoration Method Using the Accelerated and Regularized RL Algorithm[J]. Journal of Atmospheric and Environmental Optics, 2011, 6(5): 342-350.

参考文献

[1] Ayers G A, Dainty J C. Iterative blind deconvolution method and its applications [J]. Optics Letter, 1988, 13(7): 547-549.
[2] Zou Mouyan. Deconvolution and Signal Recovery [M]. Beijing: National Defence and Industry Press, 2001: 172-173(in Chinese). 邹谋炎. 反卷积和信号复原 [M]. 北京: 国防工业出版社, 2001: 172-173.
[3] Tsumuraya F, Miura N, Baba N. Iterative blind deconvolution method using Lucy's algorithm [J]. Astronomy and Astrophysical, 1994, 282(2):699-708.
[4] Richardson W H. Bayesian-based iterative method of image restoration [J]. J. Opt. Soc. Am. A, 1972, 62(1): 55-59.
[5] Lucy L B. An iterative technique for the rectification of observed distributions [J]. Astronomical Journal, 1974, 79: 745-754.
[6] Biggs D S C, Andrews M. Acceleration of iterative image restoration algorithms [J]. Applied Optics, 1997, 36(8): 1766-1775.
[7] Dey N, Blanc-Féraud L, Zimmer C. A deconvolution method for confocal microscopy with total variation regularization [J]. IEEE Biomedical Imageing, 2004, (2): 1223-1226.
[8] Biggs D S C, Andrews M. Conjugate gradient acceleration of maximum- likelihood image restoration [J]. Electronics Letters, 1995, 31(23): 1985-1986.
[9] Biggs D S C, Andrews M. Acceleration of iterative image restoration algorithms [J]. Applied Optics, 1997, 36(8): 1766-1775.
[10] Conchello J A, McNally J G. Fast regularization technique for expectation maximization algorithm for optical sectioning microscopy [C]. Three-Dimensional and Multidimensional Microscopy: Image Acquisition and Processing Ⅲ [M]. 1996, 2655: 199-208.
[11] van Kempen G M P, van Vliet L J. Background estimation in non linear image restoration [J]. Journal of Optical Society of America A, 2000, 17(3): 425-433.
[12] de Monvel J B, Scarfonec E, Le Calvez S, et al. Imgage-adaptive deconvolution for three-dimensional deep biological imaging [J]. Biophysical Journal, 2003, 85(6): 3991-4001.
[13] Dey N, Blanc-Féraud L, Zimmer C, et al. Richardson-Lucy algorithm with total variation regularization for 3D microscope deconvolution [J]. Microscopy Research and Technique, 2006, 69(4): 260-266.
[14] Biggs D S C. Accelerated iterative blind deconvolution [D]. New Zealand: Doctorial Dissertation of University of Auckland, 1998.
[15] Roggemann M C, Welsh B. Imaging Through Turbulence [M]. USA: CRC Press, 1996.
[16] Tang Xinjian. Research of Regularization Method in Image Restoration [D]. Wuhan: Doctorial Dissertation of Huazhong University of Science and Technology, 2006(in Chinese). 唐新建. 图像复原正则化方法研究 [D]. 武汉: 华中科技大学博士论文, 2006.