摘要
脉冲耦合神经网络是一种新型神经网络,该网络无需训练,根据脉冲耦合神经网络相邻神经元同步点火特性,提出了一种基于灰度特征聚类的脉冲耦合神经网络图像分割方法,利用脉冲耦合神经网络点火捕获特性,实现了对特征的自组织聚类,克服了以往基于统计方法对于相邻灰度影响的考虑,弥补了空间不连贯灰度区域分割成离散块的缺点.针对目前对网络参数的选取还主要停留在人工调整和确定阶段,对参数的选取进行了分析,并对迭代终止条件进行了研究.通过实验,证明分析结果是有效的.
Pulse coupled neural network （PCNN） is a novel architecture neural network. This neural network is based on the mammalian vision and simulates the way that the brain process visual impressions. It works without training, which means that the PCNN itself knows nothing about the features in the images. A novel image segment method is presented based on intensity clustering using PCNN, utilizing its synchronization fire characteristic to carry out the self-organization clustering. The method presented performs well in segmentation experiments; its most useful characteristic is that it can segment discontinuity area as a whole piece. Aiming at good performance, the criteria of choosing parameters are analyzed. It has been proved that the criteria is useful.
出处
《武汉大学学报（工学版）》
CAS
CSCD
北大核心
2009年第3期405-408,共4页
Engineering Journal of Wuhan University
基金
航天科技集团核心计划项目(编号:417010202)
关键词
脉冲耦合
神经网络
灰度聚类
图像分割
pulse coupled
neural network
intensity cluster
image segmentation
