中文摘要：

针对传统分水岭算法中存在的过分割现象，提出了一种修改标记符梯度图像的分水岭分割方法。该算法首先利用阈值分割法对图像中感兴趣的目标和背景进行标记；然后根据标记的二值图像，运用形态学极小值标定技术对原有梯度图像进行修正；最后，使用分水岭算法对修正的梯度图像进行分割。实验结果表明，该算法能有效地抑制过分割问题，并且使用标记符会为分割问题带来先验知识，这为人们解决模式识别等更高级的工作提供了一种有效方法。

英文摘要：

Aiming at the traditional watershed segmentation algorithm over existing phenomenon, a modified gradient image tag watershed segmentation method is proposed. Firstly ,threshold segmentation method uses the image of interest to mark the target and background; Then labeled binary image, use minimum calibration morphological technique for the original gradient image amendment;Last use watershed algorithm for improved gradient image segmentation. Experimental results show that the algorithm can effectively inhibit the oversegmentation problem, and using the tag will bring the division of a priori knowledge, which for people to solve the pattern recognition and other more advanced work provides an effective method.