基于改进的PCNN多目标图像分割算法  

Multi-target Image Segmentation Algorithm Based on Improved PCNN

作　　者：宋寅卯[1] 刘国乐[1] 

机构地区：[1]郑州轻工业学院电气信息工程学院,郑州450002

出　　处：《数据采集与处理》2009年第4期536-542,共7页Journal of Data Acquisition and Processing

摘　　要：脉冲耦合神经网络(Pulse-coupled neural network,PCNN)可有效地用于图像分割。为获得满意分割效果,PCNN需要选取适当的参数,目前其参数往往通过反复试凑确定。针对这一问题,基于改进的PCNN模型,提出结合图像灰度直方图,以最大交叉熵函数作自适应遗传算法的适应度函数,采用自适应遗传算法搜索最优门限阈值函数的图像分割算法。实验结果表明,该算法可以有效地实现多目标图像分割,且分割效果优于多阈值Ostu算法。Pulse-coupled neural network（PCNN） can be used for image segmentation. It is necessary to select the appropriate parameters to achieve satisfactory segmentation results when PCNN is applied to image segmentation. Up to now,the parameters of PCNN are always manually adjusted. To solve this problem, a multi-target image segmentation algorithm based on the improved PCNN is proposed. It combines the gray histogram of images, uses the maximal cross-entropy function as the fitness function of adaptive genetic algorithm ,and adopts adaptive genetic algorithm to search the optimal threshold function, Experimental results show that the proposed algorithm can effectively complete multi-target image segmentation, and its segmentation results are superior to that of the multi-threshold Ostu algorithm.

关 键 词：脉冲耦合神经网络 多目标图像分割 自适应遗传算法 灰度直方图 最大交叉熵 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]