基于加速阻尼Richardson-Lucy算法的湍流退化图像盲复原方法  

Turbulence-Degraded Image Blind Restoration Method Using Accelerated and Damped Richardson-Lucy Algorithm

作　　者：李勇[1] 范承玉[1] 时东锋[1] 王海涛[1] 冯晓星[1] 乔春红[1] 徐博[1] 

机构地区：[1]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

出　　处：《激光与光电子学进展》2011年第8期104-111,共8页Laser & Optoelectronics Progress

基　　金：国家863计划(A825021;A825011);中国科学院合肥物质科学研究院计算中心(0330405002-7)资助课题

摘　　要：提出了一种基于加速阻尼Richardson-Lucy(ADRL)算法的湍流退化图像盲复原方法,称为ADRL-IBD方法。在阻尼Richardson-Lucy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ADRL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的物理模型或根据观测图像来获取初始的点扩展函数(PSF),利用阈值分割技术获取图像目标的可靠支持域,在每一次迭代中,对图像施加支持域约束。模拟图像和实际湍流退化图像复原结果表明,基于Richardson-Lucy算法的IBD算法要优于基于Wiener滤波的IBD算法,并且ADRL-IBD算法具有较强的抗噪性,与RL-IBD算法相比,收敛速度更快,复原结果更好。A turbulence-degraded image blind restoration method called ADRL-IBD using the accelerated and damped Richardson-Lucy （ADRL） algorithm is proposed. The ADRL algorithm is based on damped Richardson-Lucy algorithm and vector extrapolation acceleration technique. The observed image or the physical model of long exposure atmospheric turbulence optical transfer function is used to obtain the initial point spread function （PSF）, and threshold segmentation technology is used for getting reliable support domain of the target image. The support domain restraint of image is exerted at each iteration. The restored images of simulation data and real turbulence-degraded data show that the iterative blind deconvolution （IBD） method based on Richardson-Lucy algorithm is better than the IBD method based on Wiener filtering, and ADRL-IBD. Compared with RL-IBD algorithm, it has better performance suppressing noise, greater convergence rate, and better restoration result.

关 键 词：图像复原 湍流退化图像 迭代盲目反卷积 矢量外推加速技术 阻尼Richardson-Lucy算法 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]