基于各向同性高斯的Canny边缘检测改进算法  

A novel improved method of edge detection based on isotropic Gaussian kernels

作　　者：马蒙蒙[1] 顾梅花[1] 杨婷婷[1] 谌博[1] 

机构地区：[1]西安工程大学电子信息学院,陕西西安710048

出　　处：《电子设计工程》2016年第8期131-133,137,共4页Electronic Design Engineering

基　　金：国家自然基金项目(61401347);西安工程大学博士科研启动基金项目(BS1205);全国统计科学研究重点项目(2014LZ06)

摘　　要：针对传统Canny在空间尺度和阈值选取上需人工设置,提出了一种自适应算法。由于传统Canny在邻域内计算梯度幅值时容易受到噪声影响,本文通过垂直,水平,45°,135°方向上的一阶差分模板计算梯度幅值。实验结果证明与传统Canny算子相比,该算法能够保留较多的边缘信息,同时抑制了一些虚假边缘,且计算简单,容易实现。Due to the scale σ and thresholding values can＇t be got automatically for the conventional Canny edge detection operator, we proposed the improved method based on Canny operator to make them automatical. When gradient magnitude got by the conventional Canny operator is influenced by noise, the proposed operator computed the gradient magnitude along vertical, horizontal, 45°, 135°direction. Compared with the Canny operator, the result of experiment show that the proposed operator can reserve more edge and eliminate pseudo edge. Besides, compution of it is easy and it can be reached easily.

关 键 词：各向同性高斯 CANNY算法 自适应 梯度幅值 最大类间差分法 

分 类 号：TN911.73[电子电信—通信与信息系统]