摘要
针对不同尺度的形态学结构元素具有不同的图像边缘检测效果,提出了采用不同尺度的结构元素来检测图像边缘的方法.该方法通过小尺度结构元素的膨胀来获取大尺度结构元素,然后进行加权合成来获得边缘图像.仿真实验表明,对于具有单色背景的混有不同噪声的灰度图像,随着膨胀次数的增加,检测得到的合成图像的边缘更清晰完整,细节更丰富;将该方法应用于具有期望PSNR值的噪声图像,和其他的传统边缘检测方法相比,其检测到的轮廓更加清晰,对噪声不太敏感,能够很好地提取边缘.
The detection results of image edges are different morphologically for multi-scale structural elements. A new edge detection algorithm is therefore proposed introducing the structural elements of different scales, i.e., the small-scale structural elements are dilated to become big-scale ones, then they are synthesized in a weighting way so as to form the edge image. Simulation results showed that for the blurred images mixed with different noises in concolorous background, the more the frequency of dilation, the more complete and clearer the synthesized image with more exact details. Compared with other conventional edge detection methods, the algorithm proposed has the advantages of providing clearer edge contour, and insensitivity to noise when applying it to the images with expected PSNR （peak signal-to-noise ratio） values.
出处
《东北大学学报（自然科学版）》
EI
CAS
CSCD
北大核心
2008年第4期473-476,共4页
Journal of Northeastern University(Natural Science)
基金
国家自然科学基金资助项目(60274099)
关键词
数学形态学
边缘检测
多尺度
结构元素
膨胀
mathematical morphology
edge detection
multi-scale
structural element
dilation
