用基函数神经网络实现多阈值图象分割

Multilevel Thresholding Using Basis Function Neural Networks

本文介绍了一种用基函数神经网络实现多阈值图象分割的新方法。它从函数逼近的角度研究基于灰度直方图的多阈值分割问题，提出了一种模糊反向传播学习算法，采用该算法的高斯基函数网络能够准确检测直方图中包含的子区域和它们的分布函数，而且速度很快。实验表明本文的方法在实际图象分割中是有效的。

This paper describes a nes method for multilevel thresholding based on basis function neural networks. The multilevel thresholding problem is studied from the view of fonctional approximation. A fuzzybackpropagation learning algorithm for Gaussian basis fimction network is presented, which can exactly detect thesubregions and their distribution functions in the histogram. Further more the speed is very fast. The effectovemessof this method is demonstrated in several examples.