基于局部特征的图像边缘检测

Edge detection of image on local feature

针对传统边缘检测对噪声较敏感且不能准确定位的不足,分析了图像边缘邻域像素之间的关系,提出了基于局部特征的图像边缘检测方法,建立了判断该像素点是否位于边缘上的准则函数。本方法以像素点为研究对象,检测的边缘可能为噪声点或伪边缘点,因此利用图像边缘连续性和噪声点孤立性的特点,运用形态学提取图像边缘并克服了噪声对边缘检测的影响;然后运用形态方向梯度去除伪边缘点。实验结果表明该方法能够准确地检测出图像的边缘,并克服了传统利用微分算子检测边缘的不足。

Aiming at the shortage of the traditional edge detection methods, which were sensitive to the noise comparatively and could not locate the edge exactly, the paper analyzed the relationship of the neighborhood of pixels and put forward edge detection method of the image based on local feature. Then introdued the function expression which was used to estimate whether the pixel was on the image edge or not. The method took the pixel as the research object, and the edge point to be de- tected was perhaps the noise or just the false edge point, so according to the edge＇ s continuity of the image and the isolation of noise, used the morphology to extract the image edge and overcame the noise effect on the edge detection, then used the morphological gradient to eliminate the false edge points. The experimental results show that the method can detect the image edge exactly, and overcome the shortage of traditional methods using the differential operator to detect the image edge.