摘要
结合了边缘检测与区域生长法优点的分水岭算法近年来在图像分割领域得到了越来越多的重视,但经典的分水岭算法存在过分割的问题,从而使得物体的轮廓线掩埋在无关的分水岭线中。为了克服这种过分割问题,本文介绍了一种基于图像森林化变换的分水岭算法,它是一种将图像处理问题简化为在图形中求解最短路径森林的统一有效的方法。该改进算法在预处理阶段对图像进行平滑操作并设置适当的梯度阈值,即减少了梯度图像中非重要区域的产生。实验结果表明,该方法可提供精确且封闭的区域轮廓线。
Since the watershed algorithm combines the excellences of edge detection and region growing, greater attention has been paid to in the domain of image segmentation, but image segmentation based on classical watershed algorithm always results in over- segmentation and makes contours of the objects buried in the irrelevant watershed lines. To solve the problem, a watershed algorithm based on image foresting transform （IFT） is introduced in the paper, which is a unified and efficient approach to simplify an image processing problem to a minimum- cost path forest in a graph. In the improved algorithm, the pie - processed image is smoothed and the proper gradient threshold is set to prevent the generation of insignificant regions. The experimental results show this method can provide accurate and closed region contours.
出处
《电子测量与仪器学报》
CSCD
2005年第5期58-62,共5页
Journal of Electronic Measurement and Instrumentation
基金
江苏省计算机信息处理技术重点实验室开放资助课题(编号:KJS01023)
江苏省教育厅自然科学研究计划资助项目(编号:KK0310068)
扬州大学科研基金项目(编号:U0211097)
