摘要
区域轮廓的提取是对焦炭显微图像中不同光学组分进行分类与识别的关键。由于焦炭的光学组分在不同偏光下呈现的多样性以及该图像本身的复杂性,现有的方法提取轮廓边缘存在较大的困难。文中采用一种边界加权的改进均值偏移算法对焦炭显微图像中不同组分进行聚类,准确聚类的同时较好地保留边缘信息;然后再采用双阈值法和多边形近似得到图像的连续轮廓。实验表明,该方法抗噪性强,能够有效地提取不同组分的轮廓。利用轮廓的特征信息为焦炭显微图像某些类别的分类提供了基础。
The extraction of regional boundary of coke microscopic image is the key to classify and recognize different optical components. Because the optical components in different coke polarized showed diversity, and the complexity of the image itself, the existing methods extraction boundary meats great difficulties. In this paper, use one method based on boundary weighted means/tiff clustering, it better keeps edge information. Then, adopt estimate threshold and polygonal approximation to get the continuous boundary of the image. And experiments have shown that this method can denoise with high efficiency, and can extract the boundaries of different compounds of image accurately. Make it possible to classify certain categories of coke microscopic image using the contour feature information.
出处
《计算机技术与发展》
2011年第11期120-123,共4页
Computer Technology and Development
基金
国家自然科学基金项目(50874001)
关键词
焦炭光学组织
边界加权
均值偏移
双阈值法
coke optical organization
boundary weighted
meanshift
estimate threshold
