维纳/投影限制集混合算法实现超分辨率图像复原  

Image Restoration Based on Hybrid Wiener/Projection Onto Concex Sets Algorithm

作　　者：刘洋[1] 崔继文[1] 

机构地区：[1]哈尔滨工业大学超精密光电仪器工程研究所,黑龙江哈尔滨150080

出　　处：《中国激光》2010年第3期789-794,共6页Chinese Journal of Lasers

基　　金：国家自然科学基金(50705026);国家863计划(2008AA04Z308)资助课题

摘　　要：为了抑制经典投影限制集(POCS)算法中图像复原产生的振铃效应及噪声放大的缺点,提高复原图像的分辨率,提出了一种基于维纳(Wiener)滤波和POCS混合的超分辨率图像复原算法。利用Wiener滤波的抗噪性和全局收敛性对经典POCS算法原始估计图像f0和超分辨率复原图像fi分别进行约束,同时在POCS中加以局部变量的约束。实验结果表明,新算法比原始POCS算法的峰值信噪比(PSNR)和归一化相关性测度(NCCF)均有提高,且Wiener作为整体约束的算法更具有优势,NCCF最多提高7.28%,PSNR最多提高32.35%,具有很好的图像复原结果。In order to improve the resolution of the restored image in the original context of projections onto convex sets（POCS）,and suppress the features of edge oscillating and enlargement of noise,a motion blurred restoration approach based on hybrid Wiener/POCS algorithm is proposed.Estimated original image f0 and super-resolution restored image fi of the original POCS are respectively bounded by a Wiener set,which is noise normalized and complete convergence,and three other bounded local variance sets are also incorporated into POCS.Some experiments based on restoration of image are given to demonstrate the effectiveness of the hybrid Wiener/POCS algorithm.As a result,the bounded Wiener variance set using in complete convergence is better than any other,and in the evaluation of quality of super-resolution restoration image,the normalized co-correlation function（NCCF） improves 7.28%,the peak signal-to-noise ratio（PSNR） improves 32.35% as the largest,which effectively suppress the ringing ripples and improve the resolution.

关 键 词：图像处理 图像复原 维纳/投影限制集混合算法 振铃抑制 相关性测度 

分 类 号：TN911.73[电子电信—通信与信息系统] TP391[电子电信—信息与通信工程]