基于MTM和灰色关联的椒盐噪声滤波算法-《计算机技术与发展》

文章信息/Info

Title:: Filtering Algorithm for Removal of Salt & Pepper Noise Based on MTM and Grey Relevance Theory

Author(s):: SHEN De-hai; E Xu; HOU Jian; School of Information Science and Technology,Bohai University,Jinzhou 121013,China

Keywords:: grey relevance; MTM algorithm; reference sequence; comparative sequence; salt &; pepper noise

摘要:: 为了有效去除数字图像中的椒盐噪声,结合灰色理论和MTM滤波原理,提出了一种基于MTM和灰色关联的去除椒盐噪声的滤波算法。算法采用开关滤波原则,如果滤波窗口中心点为噪声点,统计滤波窗口内非噪声点的像素值集合,以集合中像素中值M为中心,δ为阈值选取灰度区间[M-δ,M+δ],将落在区间内的像素值确定为比较序列,然后计算这些像素的中值,将中值作为参考序列,采用均值化方法对比较序列和参考序列进行无量纲化处理,利用灰色关联分析法计算各比较序列对应元素的关联系数,将它们作为对应像素的权值系数进行加权运算,将最终的加权结果作为滤波窗口的滤波输出。如果滤波窗口中心点不是噪声点,则保持该点原始像素值不变。从实验结果可以得出,该算法和标准中值滤波算法、极值中值滤波算法、MTM滤波算法相比,具有较强的抑制椒盐噪声性能,且边缘保护效果良好。

Abstract:: In order to remove the salt and pepper noise,combining the grey theory and the principle of MTM (modified trimmed mean) filtering,we propose a filtering algorithm based on MTM and grey relevance theory. The algorithm adopts the principle of switching filtering. If the filtering window center is noise point,the set of pixel values of the non-noise points in the filtering window is counted,and the gray interval [ M - δ , M +δ ] is selected with median value M of the pixel in the set as the center and δ as threshold. We take thepixel values falling in the interval as the comparison sequence,and then calculate the median value of these pixels as reference sequence.We use mean method to carry on dimensionless treatment for the comparison sequence and the reference sequence,calculating the corresponding element’s correlation coefficients of comparison sequence,which are taken as the weights of corresponding pixels. The weighted result is obtained as the filtering output. For non-noise point,keep the original value unchanged. The simulation shows that the proposed algorithm has strong ability of suppressing the salt & pepper noise and well edge protection effect compared with standard medianfiltering algorithm,the extremum median filtering algorithm and the MTM filtering algorithm.