摘要
本文提出了一种利用神经网络实现灰度形态滤波的方法。文中首先简述了形态学的基本理论，然后给出了灰度形态学理论的两种最基本运算（膨胀和腐蚀）的神经网络实现方法。网络中的连接权值为形态滤波的结构元素，按照δ学习规则，自适应地对结构元素进行学习训练。该方法计算简单，速度快，对于提高灰度形态滤波的性能有显著成效。
This paper presents a neural network of gray morphological filters. In the structure, weights of the neural network are represented as a gray structure element and trained by a learning algorithm based on δ-Learning Criterion. The results of numerical examples will be shown that the algorithm has better filtering properties than any conventional morphological operation.
出处
《系统工程与电子技术》
EI
CSCD
1999年第3期56-59,共4页
Systems Engineering and Electronics
