基于多子窗口的混合噪声滤波算法  

A Mixture Noise Filter Algorithm Based on Multiple Sub-windows

作　　者：沈德海[1] 张龙昌[1] 鄂旭[1] 侯建[1] 

机构地区：[1]渤海大学信息科学与技术学院,辽宁锦州121013

出　　处：《计算机技术与发展》2015年第6期69-72,共4页Computer Technology and Development

基　　金：辽宁省高等学校实验室项目(L2012397);博士后基金项目(2012M520158);辽宁省"百千万人才工程"资助项目(2012921058);教育科研一般项目(L2012400)

摘　　要：为了去除数字图像中混有的高斯噪声和椒盐噪声,提出了一种基于多子窗口的去混合噪声算法。算法借鉴多级中值滤波思想,先将滤波窗口划分为纵横交叉的多个子窗口。对于椒盐噪声,先统计各子窗口内非噪声点的个数,如果为奇数,求出各子窗口非噪声点的中值;如果为偶数,计算各子窗口非噪声点的均值,然后用这些值的中值替换噪声点。对于高斯噪声,算法采用多子窗口均值法进行滤波,用各子窗口均值的均值替换噪声点。实验证明,该算法对混合噪声具有较好的滤波效果,且有效地保持了图像中的细节信息。To filter out the Gaussian noise and salt and pepper noise in the digital image,a mixture noise filter method based on multiple sub-windows is presented. The algorithm references multistage median filter theory,divides the filter window into vertical and horizontal cross sub windows firstly. For salt and pepper noise,statistics the number of the non-noise points,if the number is odd,calculating the median of all non-noise points in every sub-window,if the number is even,calculating the mean of all non-noise points in every sub-window or mean of all non extreme pixels,and then replace the noise pixels with the median of above median or mean. For the Gaussian noise,algorithm uses multiple sub-windows mean filtering,applying the mean of the all means of the very sub-window to replace the noise point. Experiments show that the algorithm has the better filtering effect for mixed noise,and keep the image details effectively.

关 键 词：多子窗口 高斯噪声 多级中值滤波 椒盐噪声 混合噪声滤波算法 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]