基于塔分割和多中心模糊聚类的医学图像分割

A Medical Image Segmentation Technique Based on Pyramidal Segemtation and Fuzzy Clustering with Multi-center Clusters

本文提出一种基于塔分割和多中心模糊C均值算法结合的无监督MR图像分割方法。文中采用根标记方法对塔图像进行过分割;在塔的最底层模糊图像上应用HSC(hierarchical subtractive clustering)计算初始的聚类中心及聚类数,进而应用FCM算法合并过分割的结果。由于塔分割有效地降低了聚类样本数和HSC自动获得有效的初始聚类中心和聚类数,实验结果表明,在聚类性能不变情况下显著地减少FCM算法的运算时间,从而实现医学图像的快速分割。

A new unsupervised image segmentation technique is presented in this paper, which combines pyramid image segmentation with the modified fuzzy c-means （FCM） clustering algorithm. Each layer of the pyramid is split into a number of regions by a root labeling technique. And fuzzy c-means is then applied to merge the regions of the layer with the highest image resolution into the number of fuzzy objects. The initial cluster centers and the number of clusters for FCM are generated by using the two-level hierarchical subtractive clustering （HSC） algorithm automatically. As pyramid segmentation can reduce the number of patterns being clustered drastically by generating a region vector instead of using each image pixel, results of experiments on actual magnetic resonance （MR） image show that the computational overhead of FCM is reduced effectively, while the segmentation results are almost the same.