基于脉冲耦合神经网络的声纳图像噪声抑制  

Noise Suppression of Sonar Images Based on Pulse Coupled Neural Network

作　　者：刘晨晨[1] 李晨[2] 张之猛[3] 

机构地区：[1]石家庄铁道学院计算机与信息工程分院,河北石家庄050043 [2]邯郸学院计算机系,河北邯郸056005 [3]哈尔滨工程大学水声工程学院,黑龙江哈尔滨150001

出　　处：《计算机仿真》2008年第10期219-222,共4页Computer Simulation

摘　　要：为解决声纳图像中存在的高斯与脉冲噪声的同时去除问题,在简化的脉冲耦合神经网络(Pulse Coupled Neural Network,PCNN)模型的基础上,提出了一种结合了数学形态学与中值滤波的噪声抑制算法。首先利用PCNN输出二值图像确定噪声在图像中的位置,用数学形态学方法保持目标的完整性,然后利用中值滤波方法去除图像的脉冲噪声并对PCNN逐次迭代的输出结果进行调整去除高斯噪声。实验结果证明,方法在保持图像边缘信息的前提下,与其他方法相比获得了更好的去噪效果。For reducing the Gaussian and pulse noise in sonar images simultaneously, a noval algorithm was proposed by incorporating mathematical morphology with median filter based on the simplified Pulse Coupled Neural Net-work （PCNN）. At first, the algorithm used the output images of PCNN to confirm the positions of noise points, and used the mathematical morphology method to keep the integrity of target in the sonar images. Then the median filter was used for suppressing the pulse noise, and the iterative PCNN outputs were modified to reduce the Gaussian noise. Simulation results show that the proposed algorithm is more effective than general de -noising methods.

关 键 词：噪声抑制 脉冲耦合神经网络 数学形态学 声纳图像 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]