摘要
在颗粒图像的自动分割中,经常遇到颗粒粘连、重叠在一起的现象,需要将它们分离为单个颗粒目标.基于边界跟踪的分离算法能够克服聚堆目标连接处的凹陷比较明显并且/或者在连接处存在灰度局部最小边缘的要求,取得了较好的分离效果.但对于某些中轴方向不是垂直或水平的聚堆目标,分割仍然不理想.为此,提出了一种改进的分离算法,该算法通过自动旋转聚堆目标区域使其中轴方向成垂直或水平来实现这种情况下的分离.实验结果表明该算法进一步提高了分离效果.
In the particle image segmentation, the objects are always touched or overlapped into each other so that it is necessary to separate or split them into single ones. The separation algorithm based on boundary tracing do not need the acute angle and/or the low density contrast at the points where the objects touched and get better separation result.However, for the cluster whose orientation of principal axis is neither horizontal nor vertical, the algorithm does not work. Therefore, an improved separation algorithm is presented in the paper to separate the cluster in this case by rotating the cluster to make its orientation of principal axis horizontal or vertical. The experiment result shows that the algorithm has improved separation efficiency further.
出处
《湖南师范大学自然科学学报》
EI
CAS
北大核心
2006年第2期28-31,共4页
Journal of Natural Science of Hunan Normal University
基金
湖南省自然科学基金资助项目(No.03JJY6025)
关键词
颗粒图像
粘连
分离算法
边界跟踪
Particle image
touched
separation algorithm
boundary tracing
