脉冲耦合神经网络与最大相关准则的图像分割  

Image segmentation using PCNN and maximal correlative criterion

作　　者：聂仁灿[1] 李莎[2] 聂彩仁[3] 

机构地区：[1]云南大学信息学院通信工程系,昆明650091 [2]成都军区装备部,成都610031 [3]云南师范大学,昆明650222

出　　处：《计算机工程与应用》2009年第35期202-204,共3页Computer Engineering and Applications

基　　金：云南省自然科学基金No.2005F0010M;云南大学重点项目(No.2004Z007C);云南大学青年项目(No.2007Q024C)~~

摘　　要：基于脉冲耦合神经网络(Pulse Coupled Neutral Network,PCNN)模型,提出了一种基于Canny算法的图像预处理方法,增强了图像的模糊边缘,提高或抑制了对应神经元的相似群捕获能力,利用最大相关准则确定神经元的衰减阈值和实现神经元相似群捕获强度的控制,避免了神经元完全捕获对分割图像的过平滑,成功实现了灰度图像的自动分割。该方法得到的分割图像取得了较好的结果,体现了更多细节,与相关文献相比,大大减少了神经元参数对分割结果的影响。Based on the PCNN（Pulse Coupled Neutral Network） model,propose an image pre-processing method based on Canny algorithm to enhance the blurry edge for image,and strengthen or restrain the capacity of capturing each other for a similar cluster of corresponding neurons.Using the maximal correlative criterion to compute the attenuation threshold,control the capturing intensity each other for neurons,and avoid the excessive smoothing aroused by through capturing for segmentation image,a new image segmentation method is put forward successfully.A good image segmentation result,which shows more image segmentation de- tails and is not influenced easily by neuron parameters compared with other corrective references,is gained using the method.

关 键 词：图像分割 脉冲耦合神经网络 最大相关准则 图像预处理 捕获强度 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]