基于动态步长的医学图像聚类分割研究

Research on Medical Image Segmentation Based on Dynamic Step Length Density Clustering

针对当前基于聚类技术的医学图像分割存在的问题,提出并实现了基于密度聚类的医学图像分割方法DSLDC-MIS。该方法在DENCLUE数据组织和密度函数构造的基础上,采用最优梯度技术实现动态步长的爬山算法分割医学图像组织。实验结果表明,DSLDC-MIS能很好地实现医学图像分割,比DENCLUE有更高的时间效率,更好地控制了聚类数目,更高的一致性和对比度。

In order to overcome the problems of medical image segmentation by current clustering technology, we offer and implement a density clustering based medical image segmentation method DSLDC-MIS. On the ground of data organization and density function construction of DENCLUE, this method makes use of optimal gradient technique to implement medical image segmentation by dynamic step length of hill climbing strategy. Experiments results show that DSLDS-MIS can segment medical image very well and has better time performance, better ability of controlling clusters number and better consistency and contrast than DENCLUE.