Abstract: The watershed segmentation algorithm is very easy to appear the phenomenon of excessive partition of this malpractice, the traditional watershed algorithm is improved. This paper puts forward a new watershed method for medical image segmentation based on threshold marking. The maximum entropy threshold method is used to extract the internal marker, and watershed transform is used on internal markers to get external marker. It modifies gradient map with internal and external marker maps and uses watershed transform on the modified gradient map, and the figure is the final segmentation graph. The experiment shows that the method for inhibiting the over segmentation is effective.

Key words: CT image, maximum entropy threshold, marking, watershed segmentation

摘要： 针对分水岭分割算法极易出现过分割现象这一弊端，对传统分水岭算法进行了改进，提出了一种新的基于阈值标记法的分水岭医学图像分割方法。其中采用最大熵阈值法来提取图像内部标记，对内部标记进行分水岭变换得外部标记。用内、外标记修正原始梯度图并作分水岭变换，所得的图为最终的分割图。实验发现，该方法对于抑制过分割是有效的。

关键词: CT图像, 最大熵阈值, 标记, 分水岭分割