基于四分法噪声检测的开关中值滤波算法  

Switching median filter with quadripartition method in noise detection

作　　者：徐超[1] 陈一虎[2] 

机构地区：[1]西北工业大学应用数学系,陕西西安710072 [2]宝鸡文理学院数学系,陕西宝鸡721007

出　　处：《计算机工程与设计》2008年第18期4765-4767,共3页Computer Engineering and Design

基　　金：国家自然科学基金项目(60672135);宝鸡文理学院院级基金项目(YK073)

摘　　要：为了精确的检测出图像中的脉冲噪声并滤除,提出一种差分四分法的开关中值滤波算法。该算法对噪声检测窗口内像素按灰度值大小排序,通过差分方法划分出高、低阶信号块和高、低阶噪声块4部分。当待测像素属于高、低阶信号块时视其为信号点,否则,根据噪声块与信号块内像素比例关系确定其为噪声点或可能噪声点,若为可能噪声点,则扩展检测窗口重新检测。对于噪声点,基于其邻域噪声密度自适应的确定滤波窗口,取滤窗内信号点的中值作为滤波输出。实验证明该算法对脉冲噪声有很强的抑制作用。A switching median filter based on quadripartition noise detection is proposed to detect impulse noise precisely and denoise. The filter arranges the pixels in the detecting window which is centered around the considered pixel according to the ascending order, and sorts the pixels into four clusters--lower, higher intensity noise clusters, and lower, higher intensity signal clusters. If the considered pixel falls into the signal clusters, it is viewed as signal pixel without processed further, otherwise, it will be classified as noise or undetermined pixel according to the number of pixels in the signal and noise clusters, the window is expanded and redetection is invoked to the undetermined pixel. As to the noise pixel, the filter window is adaptively determined according to the noise density around the pixel, the noise pixel will be substituted by the median of the signals in the filter window. Experimental results show that the filter is effective to preserve impulse noise.

关 键 词：图像去噪 脉冲噪声 非线性滤波 四分法 开关中值滤波 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]