基于改进谱聚类的医学图像分割算法

A new medical image segmentation algorithm based on improved spectral clustering

为了解决基于像素难以有效分割的医学图像问题,提出一种改进谱聚类方法:一,将全局划分成具有强关联的子问题提高图像分割精度;二,传统基于欧氏距离度量的聚类容易陷入局部最优,提出流行距离构造样本相似矩阵,从而得到图像全局上的一致。最后通过对脑核磁共振图像分割验证算法的有效性。

To solve the problem of difficult to effective segmentation of medical images based on pixel, an improved spectral clustering method is proposed. Firstly, the global divide into sub-problems associated with a strong correlation to improve accuracy of image segmentation; Secondly, based on the traditional Euclidean distance metric easily fall into local optimum, proposed manifold distance constructed sample similar matrix, resulting in a consistent global image on. Finally, through by segmented the brain magnetic resonance image to validate the effectiveness of the algorithm.