Edgeflow数字图像边缘检测  

Edgeflow Digital Image Edge Detection

作　　者：宋顶利[1,2] 杨炳儒[1] 于复兴[3] 

机构地区：[1]北京科技大学信息工程学院,北京100083 [2]河北理工大学理学院,河北唐山063000 [3]河北理工大学计算机与控制学院,河北唐山063000

出　　处：《微电子学与计算机》2011年第4期81-86,共6页Microelectronics & Computer

基　　金：国家自然科学基金项目(60875029)

摘　　要：在对数字图像的处理中,边缘检测是其重要内容.常用的图像边缘检测方法,如检测梯度的最大值法,检测二阶导数的零交叉点法,统计型方法以及小波多尺度边缘检测法等,都存在难以确定合理的参数阈值的问题.由此提出了Edgeflow方法,综合了亮度、纹理和相位等各种图像特征信息,以方向相反的边缘流相遇的位置确定对象的边缘,解决了传统基于边缘的图像分割算法难以确定合理阈值的问题.论述了基于边缘流图像分割算法的原理,对该算法进行了调整,并设计出一套预测编码模型来实现.实验结果表明:Edgeflow方法参数调整少,检测效果好.In digital image processing, edge detection is an important content. Commonly used edge detection meth- ods, such as the method of detection the maximum gradient, the method of detection the second derivative of the ze- ro-crossing point, statisticabbased method and wavelet multiscale edge detection method, etc. , there is difficulty to determine the parameters of reasonable threshold. The Edgeflow method synthesizes the brightness, texture and phase characteristics of a variety of image information to meet the opposite direction of the edge of streams to deter- mine the location of the edge of an object, it solved the problem that the threshold value of the traditional image seg- mentation algorithm based on the edge is difficult to determine. This paper discusses the principle of image segmen- tation algorithm based on edge flow, adjusts the algorithm and designs a predictive coding model to achieve. The results show that： Edgeflow method adjusts parameter less and detection effect is better.

关 键 词：Edgeflow 边缘检测 统计特征 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]