摘要
图像分割技术一直都是图像处理中的难题,原位根系CT图像的分割结果影响后续的三维重建工作,选取合适的分割算法十分关键。通过对根系CT图像分割算法以及根系CT图像特点的的研究,选择阈值分割、区域生长和FCM聚类的方法对序列图像进行分割,并将区域生长方法进行优化,减少分割时计算的步骤,同时改进f cm聚类方法,将空间信息融入到FCM目标函数中。结果表明,阈值法仅对序列初始的简单图像处理效果理想,改进的区域生长法对分割目标连续的图像分割效果好,分割速度达到0.6 s。而改进的FCM虽然需要花费23 s的时间,但对不连续的分割目标分割效果明显。三种方法均体现出分割的准确度,后两者能有效提升分割的效率。适当选取以上三种分割方法,即可快速准确的完成林木幼苗CT序列图像的分割。
Image segmentation technology has always been the key of image processing. The segmentation results of situ roots CT image sequence also affect subsequent reconstruction work, so it is crucial to select the appropriate segmentation algorithm. By stud- ying the roots CT image segmentation algorithm and the features of CT image, the threshold segmentation, region growth algorithm, and FCM were selected to slice the CT Image sequence. The region growing method was optimized to reduce the segmentation steps of calculation. FCM was also improved by integrating the spatial information into the FCM objective function. The results showed that the threshold method was only good for the initial simple image. The improved region growth algorithm was good for continuous image seg- mentation and the segmentation speed could reach 0. 6s. The improved FCM costs 23s, however it is obviously good at the discontinu- ous segmentation for object segmentation. The three methods can all achieve the accuracy of segmentation, and the latter two even en- hance the efficiency of segmentation. Therefore, these three segmentation methods should be selected appropriately in order to quickly and accurately segment the CT image sequences of the seedling roots.
出处
《森林工程》
2014年第1期25-29,共5页
Forest Engineering
基金
中央高校基本科研业务费专项资金(DL11BB33)
国家自然科学基金(31270757)
博士学科点专项科研基金(20120062120008)
关键词
植物根系
图像分割
区域生长
FCM
plant root
image segmentation
region growth
fuzzy C means
