一种用神经网络来进行图像恢复的方法

本文首先简要地介绍了Ｙ．－Ｔ．Ｚｈｏｕ提出的图像恢复神经网络方法，并分析了这种方法存在的缺陷，在此基础上提出了一种用于图像恢复的改进神经网络方法，在计算机上利用实际拍摄的人头图像对本文提出的改进方法进行仿真实验，结果表明本文提出的这种方法是切实可行的。

A Neural Network Algorithm for Image Restoration

In this paper,the problem of image restoration neural network is discussed with.At first, some typical algorithms for image restoration are reviewed in brief. Then the neural network proposed by Y.-T. Zhou(ZNN)for image restoration is introduced in detail. The problems of ZNN are analysed,and a modified neural network (MNN) is proposed for image restoration. This MNN has less complexity in time-and space-domain than that of ZNN and thereby it can operate more quickly.Simulations are run for restoration of blurred and degenerated girl's photograph using MNN,ZNN,inverse filter,and Wiener filter, respectively. The corresponding results indicate that:i)The image restorated by MNN or ZNN is obviously better than that by inverse filter; ii) The image restorated by MNN or ZNN is smoother than that by Wiener filter;and iii) The image restorated by MNN is about the same quality as that by ZNN,but MNN operates more quickly than ZNN.