基于多尺度形态学标记的分水岭算法

Watershed Algorithm Based on Multi-Scale Morphological Marker

为了解决传统分水岭的过分割问题,提高图像分割的准确性,本文提出一种多尺度形态学标记的分水岭算法。首先利用H-minima变换对中值滤波后的梯度图像进行初始标记;然后利用不同尺寸的结构元素分别进行梯度图像闭重构,对闭重构后的各梯度图像与梯度图像求差,得到标记图像,将各尺度下的区域标记点求并集,并作为新的标记图像,获得最终的标记;最后对修改后的梯度图像进行分水岭分割。实验结果表明,该方法能有效的解决过分割问题,边界定位准确。

In order to solve the over-segmentation problem of traditional watershed and improve the accuracy of image segmentation, this paper proposes a watershed algorithm based on multi-scale morphological marker. Firstly we use the H-minima transformation to preprocess the gradient image which is proposed by the median fil-tering. Then we perform the closing reconstruction of gradient images by using the structural elements of different sizes and we obtain the differences with the gradient image which are considered as marker images. The union set of a series of marker images are regarded as the final mark; Finally, the watershed transformation of modified gra-dient is performed. The experimental results show that this method can effectively solve the problem of over-seg-mentation and can define the boundary precisely.