基于形态学分水岭的Normalized Cut图像分割方法

针对传统分水岭分割后所产生的过分割问题，提出了一种基于形态学分水岭算法和Normalized Cut算法相结合的图像分割方法。在传统分水岭分割的基础上，融合形态学算法进行初步分割，并将分割后各个子区域的形心和平均灰度值作为Normalized Cut算法的输入参数，完成图像分割。结果表明，组合算法既避免了过分割现象，也达到了Normalized Cut算法的分割精度，是一种有效的图像分割算法。

A Normalized Cut Image Segmentation Method Based on Morphological Watersheds

In view of the over-segmentation caused by traditional watershed segmentation, a normalized cut image segmentation method combined on morphological watershed transformation is proposed in this paper. On the basis of traditional watershed segmentation, the morphology algorithm is adopted for initial segmentation, and the average grey value of each segmented area is used as input parameters for the normalized cut algorithm to complete the image segmentation. The results show that combined algorithm not only avoids over-segmentation, but also achieves as good accuracy as the normalized cut algorithm segmentation.