摘要
提出了一种用概率神经网络(PNN)检测图像随机脉冲噪声点方法。首先提取已知图像脉冲噪声像素点的特征作为PNN的输入,然后建立了PNN脉冲噪声点识别模型,再对其它噪声图像的每一个像点进行识别,最后只对噪声点进行中值滤波。Matlab仿真实验表明,同BPNN检测方法相比,该网络能明显提高识别正确率,因此有更好的脉冲噪声滤除效果,且该方法滤除脉冲噪声简单快速,是一种较好的神经网络图像脉冲噪声识别滤除方法。
Probabilistic Neural Network（PNN） is proposed for random impulsive noise detection in this paper. Firstly, the characters of a known image pixels with impulsive noise are distilled as the PNN＇s input, then, PNN impulsive noise detection model is obtained, and finally, only the noisy pixels of other noisy image detected by PNN are processed by median filter. The result by Matlab shows that this neural network, compared with BPNN, can remarkably raise the correct noise identification rate, thus resulting in better impulse noise elimination. Besides, being simple and fast, this method proves to be a promising neural network impulsive noise detection and suppression method.
出处
《电讯技术》
北大核心
2009年第12期35-38,共4页
Telecommunication Engineering
基金
国家自然科学基金资助项目(60573040)
关键词
图像处理
概率神经网络
脉冲噪声
噪声检测
去噪
中值滤波
image processing
probabilistic neural network （PNN）
impulsive noise
noise detection
noise suppression
median filter
