基于标记分水岭和FLICM模糊聚类的图像分割方法研究

Image segmentation based on marker-based watershed and fuzzy clustering of FLICM

传统的分水岭算法的应用非常广泛,但是存在过分割的问题。通常有两类方法解决该问题。第一类是后处理方法,它的原理是根据分水岭分割后的结果,使用某种方法让一些区域合并在一起。第二类属于前处理方法,在应用传统分水岭算法之前先标记提取,目前已经提出了基于标记的分水岭分割算法。这种方法虽然可以在一定程度上缓解传统分水岭算法的过分割问题,但是还是会有一定的过分割。文章在基于标记的分水岭算法的基础上,利用局部信息模糊C均值聚类算法(Fuzzy Local Information C-Means Clustering,FLICM)进行区域合并。实验结果表明:所提出的方法能有效地解决图像过分割问题,且更趋近于自然分割。

The traditional watershed algorithm is widely used,but with this algorithm,it is difficult to avoid over-segmentation. There are usually two kinds of ways to solve the problem. The first one is post-treatment method,which uses some methods to make regional merging on the results of watershed segmentation. The second one belongs to the pre-treatment method,which makes the marker extraction before applying the traditional watershed algorithm. The method of marker-based watershed has already been proposed. To some extent,this method can solve the problem of over-segmentation in traditional watershed algorithm,but there will still be some over-segmentation. Based on the marker-based watershed algorithm,this paper uses Fuzzy Local Information C-Means Clustering,FLICM,to make regional merging. The experimental results show that the proposed method can effectively solve the problem of over-segmentation in images,and is closer to natural segmentation.