噪声概率快速估计的自适应椒盐噪声消除算法  

Adaptive removal of salt-pepper noises through fast noise ratio estimation

作　　者：徐冠雷[1] 王孝通[2] 徐晓刚[2] 朱涛[1] 

机构地区：[1]大连舰艇学院博士生队,辽宁大连116018 [2]大连舰艇学院航海系,辽宁大连116018

出　　处：《光电工程》2005年第12期34-38,共5页Opto-Electronic Engineering

基　　金：国家自然科学基金资助项目(60473141)

摘　　要：提出一种可识别噪声概率自动调节滤波窗口的自适应椒盐噪声消除算法。对非理想椒盐噪声污染图像随机区域进行变窗口中值滤波,将结果与滤波前比对获得噪声点数,滤波区域即按此点数排序。然后取每种滤波窗口下的中间三组数据,该数据平均加权获取图像噪声概率初估计,对初估计平均加权即得图像噪声概率。滤波前首先采用阈值法排除明显噪声点,剩余像素中再以离窗口中心像素距离平方的倒数为权值估计中心像素。最后由噪声概率按照T-S模糊规则对不同模型的输出估计值进行融合。实验证明,与传统中值滤波等算法相比,该算法具有噪声自动估计和自适应窗口调节能力,滤波后标准均方差可减少20%以上,速度可提高一倍多。An adaptive salt-pepper removal algorithm is proposed, which can estimate noise ratio and change the filtering window automatically. In non-ideal image with salt-pepper noises, some regions are chosen at random from which every pixel is filtered by median filter through different filtering windows. Numbers of noise pixels for every region are obtained by comparing the filtered regions before filtering. These numbers are ranked and the median three numbers are taken to be weighted to get a ratio for every kind of window, thus three ratios are got and they are weighted to sum up to achieve the noise ratio. Then through T-S fuzzy model, choose the filtering model and the max filtering window and exclude noise pixels by way of thresholding in the filtering window. The rest pixels are summed up with adaptive weight according to the distance from centered pixel to get the estimation of the center pixel. Experiments prove that the proposed algorithm has higher accuracy with NMSE 20% smaller and is one time faster than classical median filter and center weighted median filters etc.

关 键 词：图像处理 噪声图像 中值滤波 椒盐噪声 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]