脉冲耦合神经网络(PCNN)模型在数字图像处理中有着广泛应用,但基本都是从网络的外在特性出发并结合一定的实际应用对其进行研究和改进,缺乏对模型本身数学特性的分析.本文从PCNN模型的迭代方程出发,对无耦合连接和耦合连接两种状态下的PCNN数学模型进行了点火机理分析,揭示了PCNN模型本身的数学耦合特性(点火阶梯)以及其对网络生物学特性(脉冲发放特性)会造成干扰和影响的现象,并分析了这种干扰和影响产生的机理和消除方法,同时提出PCNN用于图像分割时参数自适应设定的方法.最后给出了更能体现神经网络生物学特性的PCNN改进模型,将其用于Lena等图像的分割处理中,取得了良好的效果.
Pulse Coupled Neural Network (PCNN) model has been widely used in digital image processing,but most of the study and the improvement of the model are based on the external characteristic of PCNN, ignoring the shortage resulted from the discretization of the neuron firing time. For both the non-linking PCNN and coupled linking PCNN, this paper analyzed the firing mechanism of neurons from the point of PCNN iterative equations, and the modified formulas of the neuron firing time are pro- posed. The phenomenon that the mathematics coupled characteristic of model itself has an overwritten effect on its biological charac- teristics is revealed. On that basis, the method of eliminating the mathematics coupled characteristic is proposed and the automatic parameters determination of PCNN is implemented. Then, a new modified PCNN model is proposed, and it is used in image segmen- tation and a ~,ood result is obtained.