中文摘要：

准确快速地分割CT切片特征轮廓是医学图像三维重建的重要环节.现有的轮廓分割方法必须通过手动层层交互操作,不仅耗时而且分割精度不高.针对这种局限性,提出一种基于启发式牙颌CT影像自动分割方法.首先用拉普拉斯算子对CT图像序列进行边缘增强,其次用轮廓匹配映射技术实现轮廓启发式传递,最后基于收缩包围算法自动分割牙颌序列.以14例完整牙（每例28 ～ 32颗牙数据样本）锥束CT断层扫描图像序列进行实验,在相同条件下分别用所提出的轮廓自动提取方法和其他提取方法,对实验样本进行轮廓提取,得到单颗牙轮廓提取的平均用时和提取轮廓与真实轮廓之间的距离差平均值.实验结果显示,轮廓自动分割算法提取单颗牙轮廓的用时约为其他手工分割法提取单颗牙轮廓用时的23％,同时提取的轮廓质量和用传统方法提取的轮廓质量相当.该方法为CT数据特征区自动化分割提供一种可行且高效的方法,为进一步改进现有的CT影像分割和三维重建算法提供了新的思路.

英文摘要：

Segmenting CT slices featured contour accurately and rapidly is an important part in the medical image 3D reconstruction. The existing contour extracting technologies have to manually extract outlines interactively layer by layer, not only time-consuming but also low accuracy. For this limitation, this paper proposed a automatic contour extracting method for tooth segmentation based on heuristic in dental and maxillofacial CT images. First, we enhanced all the image edges with Laplace operator. Second, the contour registration mapping technology passes the outline down heuristically. Ultimately, the shrink-wrapping algorithm is used to segment the teeth outlines automatically. With 14 cases （28 -32 teeth samples） of a maxillary tooth cone beam CT tomography image sequences, at the same situation, we extract the contours of the sample with our method and the traditional method respectively. We determined the time of each tooth and the average distance between extracted contour and the real contour. From the experimental result, we concluded that the average time of our automatic extracting algorithm was about 23% of the traditional interactively algorithm time. The quality with our method is similar to traditional extracted contour. This method can provide a feasible and efficient approach to automatically segment CT slices featured areas, and further provide a new idea to improve the contour automatic extracting and image 3D reconstruction algorithm.