基于改进多级中值滤波的加权滤波算法  

A weighted filtering algorithm based on improved multilevel median filtering

作　　者：沈德海[1] 侯建[1] 鄂旭[1] 张龙昌[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：《现代电子技术》2015年第12期85-88,共4页Modern Electronics Technique

基　　金：国家自然科学基金项目:基于博弈论的高效稳定聚类算法研究(61473045);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);教育厅科研一般项目(L2014451);辽宁省社会科学规划基金项目(L14AGL001)

摘　　要：为了有效地抑制图像中的椒盐噪声,更好地保持图像细节,提出一种基于多级中值滤波的加权滤波算法。算法采用5×5滤波窗口,如果中心点为噪声点,则将滤波窗口划分为水平和垂直10个条形子窗口,先计算每个子窗口内所有非噪声点的均值,作为加权运算的基础值,然后求出这些基础值的中值,利用每个基础值与它们中值的差计算出每个基础值的相应权值。最后将这些基础值与对应权值进行加权运算,将结果替换中心点的像素值;如果中心点为非噪声点,则保持原值不变。实验结果表明,该算法对于高密度椒盐噪声污染的图像具有良好的去噪性能,并且较好地保持了图像的细节,效果优于传统的中值滤波算法和多级中值滤波算法。In order to suppress the salt and pepper noise in image effectively,and preserve the image details well,a weighted filtering algorithm based on multilevel median filtering is proposed. The 5×5 filtering window is used in the algorithm, but if the central pixel is noise point,the filtering window is divided into vertical and horizontal ten striptype subwindows. The mean value of all nonnoise points in each subwindow is computed as the basic value of weighting operation,the median of these basic values is found out,and then the corresponding weight of each basic value is worked out according to the differences between these basic values and their median. The weighting operation of basic values and the corresponding weights is conducted. The result is used to replace the pixel value of central point. If the central point is nonnoise point,the initial value is kept. The experiment results indicate that the proposed algorithm has good denoising performance for images polluted by high density salt and pepper noise,at the same time,keeps the image details well. Its effect is superior to the traditional median filtering algorithm and multilevel median filtering algorithm.

关 键 词：多级中值滤波 椒盐噪声 条形子窗口 加权滤波算法 

分 类 号：TN911.73[电子电信—通信与信息系统] TP391.41[电子电信—信息与通信工程]