融合边缘信息的极值中值滤波算法  

Algorithm of Extremum Median Filtering Fusing Edge Information

作　　者：金秋春[1] 童小利[2] 薄树奎[1] 

机构地区：[1]郑州航空工业管理学院计算机系,河南郑州450015 [2]郑州航空工业管理学院机电工程学院,河南郑州450015

出　　处：《机床与液压》2013年第3期38-40,共3页Machine Tool & Hydraulics

基　　金：国家自然科学基金项目(41001235);河南省科技厅科技公关项目(102102210443);河南省教育厅自然科学研究计划项目(2011A210025)

摘　　要：针对标准中值滤波算法对所有像素进行统一处理造成图像模糊的不足,提出融合边缘信息的极值中值滤波算法。该方法首先利用噪声与邻域内其他像素无相关性、而图像边缘像素在边缘方向上具有显著相关性这一特点检测噪声,生成噪声标记图;然后根据噪声标记图只对噪声像素采用邻域内非噪声值的中值滤波。在不同类型和不同强度噪声图像上进行了多种方法的对比实验与分析,结果表明:融合边缘信息的极值中值滤波算法在去除噪声的同时能更好地保护边缘细节信息,特别是噪声密度较大时(D>0.5),性能明显好于极值中值滤波。According to the deficiency of standard median filter, an algorithm of extremum median filtering fusing edge informa- tion was proposed. Noise pixels were detected based on the correlation characteristic of image pixels and flag map of noise pixels was produced. Then only the noise pixels were processed by the median value of neighborhood non-noise pixels. Through contrast test and analysis on the different type and different noise intensity images, the proposed algorithm is effective in denoising and edge preserving. Especially, when the noise intensity exceeds 50%, it performs better than EM filter.

关 键 词：噪声检测 图像边缘信息 极值中值滤波 信噪比 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]