基于NAS-RIF算法和神经网络的图像盲复原  

Blind Image Restoration Algorithm Based on NAS-RIF Algorithm and Neural Network

作　　者：强俊[1] 楚宁[1] 周鸣争[1] 汪军[1] 

机构地区：[1]安徽工程科技学院计算机科学与工程系,安徽芜湖241000

出　　处：《计算机技术与发展》2007年第6期104-106,共3页Computer Technology and Development

基　　金：安徽省教育厅自然科学基金重点资助项目(2004kj033dz)

摘　　要：在图像盲复原中,NAS-RIF算法在无噪情况下,能够得到较好的复原结果,但是对有观测噪声的图像复原效果不理想。而Hopfield神经网络有利于缓解图像复原过程中的震铃效应,但前提是知道退化图像的点扩展函数。将二者相结合提出一种基于NAS-RIF算法和神经网络的图像盲复原新算法,首先由NAS-RIF算法先估计出退化图像的点扩展函数,再利用Hopfield神经网络算法对其进行复原。实验结果表明,该算法具有较好的盲复原效果。In blind image restoration, the NAS- RIF algorithm can obtain the better restoration result while without noise. When observational noise exists, the restoration obtained by NAS- RIF algorithm isn＇t ideal, However, Hopfield neural network is good at alleviating ringing effect in the process of restoration. But the premise is that the PSF of the degraded image has been known. Gives a new blind image restoration algorithm which based on NAS- RIF algorithm and neural network. NAS - RIF algorithm estimates PSF from degraded image at first , and then neural network restores. The result of the experiment shows that this algorithm will produce better effect than others.

关 键 词：图像盲复原 NAS-RIF HOPFIELD神经网络 

分 类 号：TN911.73[电子电信—通信与信息系统]