脉冲耦合神经网络(Pulse Coupled Neural Network)是基于动物视觉图像形成机制, 用一组数学式表达这种机制的仿生学方法。PCNN的数学表达式中有 7个关键的参数, 而其中的阈值放大系数 VE决定了 PCNN网络中每个像素的分割阈值大小。通过平均阈值算法和 Ostu算法分别计算出图像的分割阈值, 并基于高斯分布模型用数学方法证明了在最小交叉熵时的最佳分割阈值在这 2个阈值构成的区间内, 通过在这 2个阈值构成的区间内搜索新的阈值作为 PCNN的参数 VE的值, 并将此寻优的参数 VE代入改进的 PCNN算法进行图像分割。在计算机上进行仿真实验, 与基于经验值的指数衰减算法比较, 该文算法分割出的兴趣区域清晰、准确, 边缘连接性好, 信息全面, 算法的效率更高, 具有很好的实用性。

The Pulse Coupled Neural Network is based on the research of animals’ visual image formation system, which bionics method is presented by a group of mathematics formulas. The formula group of PCNN has 7 key parameters, among which the factor of threshold amplification VE decides the threshold of segmentation of the image and the output of bilinear image. This article calculates the threshold by the mean threshold and Ostu method, then proves that at the minimum cross-entropy, the best threshold is in the interval of the two thresholds by Gauss distribution model, finally searches a new threshold as the value of PCNN parameter VE between these two thresholds. We divided the image by using the changed PCNN, simulate the algorithm on the computer and compare the result with experiential decay method. The result shows that our algorithm divides a clear and accurate interesting area, the connectivity is better, the efficiency is better, so this algorithm has a good practicability.