利用灰色理论构造统计量进行图像边缘检测

提出了基于灰色绝对关联度构造统计量进行图像边缘检测的新算法。所提算法将像素8邻域按水平、垂直方向分为4个部分，再按两对角线方向分为另外4个部分，基于灰色绝对关联度模型构造一种统计量，分别计算8个部分内像素的统计值，取最小的统计值与阈值比较，从而确定邻域中心像素是否为边缘点及其边缘方向。实验表明，所提算法简单、易实现，具有一定抗噪性能及检测边缘准确清晰等优点，对于灰度变化剧烈的图像检测出的边缘效果比传统算法要好。

Statistic-based image edge detection about gray system theory

A novel algorithm to detect edges on gray images based on a statistic is proposed, which is referred to grey absolute correlation degree. First, eight-neighborhood of a pixel is divided along horizontal, vertical and two diagonal direction, respectively. Then the statistic value for each part of the pixels is computed. Finally edge point and edge direction are judged based on the comparison of the minimum statistic value with the threshold value. For grayscale variation acutely images, the experimental results show that the proposed algorithm not only is of advantages of simple, easy to be realized, certain anti-noise ability, but also can gain more precise edge pixel than several classical edge detection algorithms.