结合灰度化权值调整的改进PCNN图像分割  

Improved PCNN Image Segmentation Combined with Grayscale Weight Adjustment

作　　者：马跃辉 辛月兰 杨晨晓 MA Yuehui;XIN Yuelan;YANG Chenxiao(College of Physics & Electronic Information Engineering,Qinghai Normal University,Xining 810000)

机构地区：[1]青海师范大学物理与电子信息工程学院,西宁810000

出　　处：《计算机与数字工程》2019年第4期924-929,979,共7页Computer & Digital Engineering

基　　金：国家自然科学基金项目(编号:61662062);青海省自然科学基金项目(编号:2016-ZJ-745)资助

摘　　要：针对传统脉冲耦合神经网络在图像分割中存在目标和背景混叠且分割时间过长的问题,提出了一种结合灰度权值调整的改进PCNN图像分割方法。该方法利用彩色图像R、G、B三分量作为PCNN的输入信息且采用双阈值的判别模式找到PCNN的初始动态阈值以减少迭代次数;然后,根据不同通道分割目标和背景的差异大小改变输出权值的比例以得到最佳分割结果。实验结果表明,论文方法可以取得较优的分割结果,与传统PCNN方法相比节省了时间,提高了分割效率。Aiming at the problem that the incomplete partition due to the intensity overlap between background and human targets,a modified pulse coupled neural networks method with weight adjustment for color-to-gray is proposed.This method uses RGB color information as input information of PCNN and the initial dynamic threshold of PCNN is found by using the model of binary threshold value to reduce the number of iterations.Then,the proportion of the output weight is according to the difference between the segmentation targets and the background to obtain the best segmentation results.Experimental results show that a modified pulse coupled neural networks method based on weight adjustment for color-to-gray can obtain a better segmentation result.Moreover,this method can save time and improve segmentation efficiency compared with the traditional PCNN method.

关 键 词：图像分割 脉冲耦合神经网络 灰度化 彩色模型 

分 类 号：TP751[自动化与计算机技术—检测技术与自动化装置]