一类用神经网络实现图像去噪的新方法

北京大学学报（自然科学版）

一类用神经网络实现图像去噪的新方法

汪海明¹, 余道衡¹, 郭仕德²

¹北京大学信息科学技术学院，北京，100871; ²北京大学遥感与地理信息系统研究所，北京，100871，Email: Wang_haiming@263.net

收稿日期:2002-09-10 出版日期:2003-09-20 发布日期:2003-09-20

A Class of New Approach for Image Restoration Based on CNN

WANG Haiming^１, YU Daoheng¹, GUO Shide^２

^１Department of Electronics, The national laboratory of machine perception; ^２The institute of remote sensing and GIS, Peking University, Beijing, 100871, E-mail: Wang-haiming@263.net

Received:2002-09-10 Online:2003-09-20 Published:2003-09-20

摘要/Abstract

摘要： 提出了一类基于二维细胞自动机(CA)和细胞神经网络(CNN)的图像滤波新方法。通过对两种新的二维CA规则的深入研究，得出了两类CA演化规则能够用于设计新的细胞神经网络，可以实现灰度图像的噪声消除。核心思想是用两个对偶CA规则设计两个CNN网络，联合实现图像滤波处理，能够获得比传统算法好得多的处理速度和效果。仿真结果证明本文的想法是合理的，期望能够在二维CA研究与应用设计方面有所启发和突破。

关键词: 细胞自动机(CA), 细胞神经网络(CNN), 图像去噪, 联合处理

Abstract: A new class of approach for image restoration based on two-dimensional Cellular Automata (CA) is proposed. By researching two new kinds of two-dimensional CA, we find two new classes of CA rules to design CNN, which can implement image restoration. The main idea is to design two CNNs with two antithetic CA rules. By this way, Image restoration can be implemented, getting obviously better results than the traditional ones. The simulation results prove our idea reasonable. It is expected to have profound influence on the technology of two-dimensional CA.

Key words: cellular automata(CA), cellular neural network(CNN), image restoration, joint processing

中图分类号:

TN911.73

汪海明, 余道衡, 郭仕德. 一类用神经网络实现图像去噪的新方法[J]. 北京大学学报（自然科学版）.

WANG Haiming,YU Daoheng,GUO Shide. A Class of New Approach for Image Restoration Based on CNN[J]. Acta Scientiarum Naturalium Universitatis Pekinensis.

导出引用管理器 EndNote|Ris|BibTeX

链接本文: https://xbna.pku.edu.cn/CN/

https://xbna.pku.edu.cn/CN/Y2003/V39/I5/666