受噪声污染图象中的基于排序统计的非线性边缘检测器

Nonlinear Rank-Order=Based Edge Detectors for Noisy Images

本文主要研究了如何对受混合噪声污染的图象进行边缘检测，并提出了一种基于排序统计的非线性边缘检测算子。该算子将输入样点集划分成两个具有不同灰度值的子集，通过子集的运算值之差判断边缘是否存在。子集的划分和输出形式的选择是减少噪声对边缘检测器影响的关键。计算机模拟实验表明，基于排序统计的边缘检测算子能较好地同时消除高斯噪声和脉冲噪声对边缘检测器的影响，并且能精细地提取图象的边缘，效果优于其它边缘检测算子。

This paper mainly studies how to detect the edges of noisy images corrupted by kinds of noises. A nonlinear rank-order-based edge detector is presented. This operator divides the input sample set into two subsets which have different grey levels and decides to by claculating the differs of the outputs of the two subsets. The division of Subsets and the seleCtion of output function is of great importance to reduce the effects of noises on the edge detector. It is showed by computer simulations that the influence of Gaussian noise and im pulsive noise on edge detecting will be eliminated by the rank-ode-bud edge detector at one time and the operator can detect fine be of hages in the mean time. The lank -order-edge edge detector will outperform other edge detectors.