基于运动估计的交互式医学图像序列分割  

Motion estimation based interactive medical image series segmentation

作　　者：胡永祥[1] 蒋鸿[1] 

机构地区：[1]湖南工业大学计算机科学与技术系,湖南株洲412008

出　　处：《计算机工程与设计》2007年第5期1098-1099,1231,共3页Computer Engineering and Design

摘　　要：针对医学图像难以自动分割,而医学图像序列采用手工分割时工作量巨大、效率低的问题,提出了一种新的交互式图像序列分割方法。在计算机的辅助下,用手工精确地描画出第一幅图像中对象的边界轮廓。后续图像的分割曲线用运动估计的方法自动得到。每完成一幅图像的分割用户都可以检查分割效果,如果不满意则可用手工修正。这个过程重复进行,直到整个图像序列分割完毕。实验结果表明,该方法能精确、快速地实现医学图像序列的分割。In order to resolve the problems that medical image is very difficult to be segmented automatically and it is very inefficient to segment medical image series by manual, a new interactive image series segmentation method is proposed. Initially, the object contour in the first image is specified interactively using a computer-aided method. The images followed are automatically segmented using motion estimation. After each image segmented, there is a pause and user may check the result. If user is not satisfied with the result, he can amend it manually. This process iterates until all images are well segmented. Experiments show that the proposed method achieves the result accurately and rapidly.

关 键 词：医学图像序列 图像分割 运动估计 交互 三维重建 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]