摘要
常用的微分边缘检测算法往往无法设立合适的阈值将影像中梯度较小的模糊边缘检测出来。针对这一点,本文提出了两种解决的方法:将图像方差标准化,拉大模糊边缘处的梯度值,或者通过设置sigmoid函数,将像素点所在区域的信息传递到梯度值中去,对梯度值进行调整,这样就能够设定合适的阈值,有效地将模糊边缘提取出来。本文将这两种算法和常用的一些微分边缘检测算法比如Sobel、LOG算法进行了比较。试验表明,这两种方法都能够有效地提取出模糊边缘。
The traditional difference edge detectors can＇t set suitable threshold for fuzzy edge detection when the gradient is small. In this paper, two methods are proposed to enhance the difference detector： enlarging the image gradient by using local image variance normalization preprocessing, and adjusting the gradient by setting sigmoid function to transfer some region information to image gradient. Accordingly, the appropriate threshold can be set for fuzzy edge detection. The proposed algorithms are compared with conventional methods such as Sobel and LOG edge detectors. Results show that the two new methods can effectively extract the fuzzy edges.
出处
《光电工程》
EI
CAS
CSCD
北大核心
2007年第7期141-144,共4页
Opto-Electronic Engineering
基金
国家自然科学基金资助项目(NSFC40371087
NSFC40401042)
中国科学院知识创新工程重要方向项目(KZCX3-SW-338)
关键词
边缘检测
方差标准化
SOBEL
遥感图像
edge detection
variance normalization
Sobel
remote sensing images
