区域差异性测度选点向量中值滤波  

Effectively Filtering Corrupted Near Bank Image Based on Discrepancy Measure

作　　者：邵承会[1] 唐可洪[1] 朱黎辉[1] 

机构地区：[1]吉林大学机械科学与工程学院,吉林长春130025

出　　处：《光电子．激光》2007年第12期1449-1452,共4页Journal of Optoelectronics·Laser

基　　金：国家"863"计划资助项目(2002AA423150)

摘　　要：提出区域差异性测度(DM),它反映一个像素点与它周围像素点的不协调程度,由此判断该点是否为噪声点。优化了DM的2个主要参数,并以此为基础提出了DM选点矢量中值滤波(DMNVM),即用DM判断移动窗口中心像素点是否为噪声点,若是利用矢量中值滤波(VMF)处理,否则保持原像素点不变。结果表明,DMNVM滤波性能明显优于现有的VMF、递进开关中值滤波(PSM)、标量中值滤波(SMF)、与标量中值距离最小向量滤波(MMVM)和方向矢量中值滤波(VDF);同近年提出的自适应向量滤波(AVMF)比较,AVMF客观指标平均绝对误差(MAE)、均方误差(MSE)和标准色差(NCD)分别是DMNVM的1.81、2.59和2.17倍,这说明DMNVM的滤波性能有显著提高。A new filter-DMNVM is proposed based on discrepancy measurement（DM）. First,it is estimated to determine whether the moving window center pixel is a corrupted pixel. If the central pixel is corrupted,it will be processed by vector median filter（VMF） ;and if the central sample is noise-free,it remains unchanged. The performance of the proposed method is compared with that of the well-known vector standard filters such as the VMF,progressive switch median filter（PSM）, scalar median filter（SMF） ,minimum distance between central sample and scalar median in its neighborhood vector median filter（MMVM） and the basic vector directional filter（VDF）. The results show the proposed method is of excellent signaldetail preservation and effective impulse noise suppression. Moreover the values of objective criteria--mean absolute error （MAE） ,mean square error（MSE） ,and normalized color difference（NCD）of DMNVM are 0.81,1.59,1.17 times less than that of adaptive vector median filtering（AVMF） respectively.

关 键 词：滤波 向量中值 区域差异性测度 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]