基于二级检测的椒盐噪声滤除算法

摘要

针对目前算法不能适用于含不同密度椒盐噪声的图像去噪,提出了一种基于噪声点极值特性和方向差异性的开关型噪声滤波器.检测阶段,采用极值法检测噪声,并计算噪声密度,当噪声密度低于阈值时进行基于噪声点方向差异性的二次检测,否则直接进行滤波；滤波阶段,根据局部噪声密度选择递归或非递归窗口,进行除去窗内噪声点的加权均值滤波.实验结果表明本文算法滤波性能和细节保护都优于传统中值滤波和一些改进算法.%As the present algorithms are not suitable to be applied to denoise images containing different density salt and pepper noise,a switching filter based on noise's extreme value and directional difference is proposed. In noise detection stage,it adopts the Minimum-maximum noise detector and calculates the noise density. Then it detects the noise for the second time based on the noise's directional difference if the density is lower than the threshold, otherwise direct filtering is made. In filtering stage,the proposed algorithm selects a recursive or non-recursive window according to the partial noise density and realizes the weighted mean filtering for removing the noise. Experimental results indicate that the filtering performance and edge-preserving of the proposed algorithm are superior to the traditional median filer and some improved algorithm.