基于形态学标记连通的分水岭图像分割

Watershed segmentation based on morphological marker-connection

为了解决传统分水岭算法的过分割问题,本文提出一种基于H-minima标记连通的图像分割的分水岭算法。该算法首先利用形态学中的H-minima方法对预处理后的梯度图像进行初始标记,然后利用连通域标记的方法进行重标记,利用强制最小标定技术使极小值只出现在标定的位置,最后,对标记后的梯度图像进行分水岭算法的图像分割。实验结果表明,该算法既能有效的解决分水岭算法的过分割问题,又能很好的保留图像中物体的边缘。

In order to solve the problem of over-segmentation in traditional watershed algorithm, a markerconnected watershed algorithm based on H-minima is proposed in the paper. The algorithm firstly uses H-minima in morphology to mark the gradient image which is pretreated. Then, it uses the marker-connected method to remark the gradient image and the minima imposition technology to make the minimum points only appear in marked position. Finally, it makes use of the watershed algorithm to segment the marked gradient image. The experimental results show that this method can not only solve the problem of over-segmentation effectively, but also reserve the edge in the image well.