基于图模型的医学图像聚类算法

医学图像聚类算法的研究是面向特殊领域图像挖掘的重要组成部分，由于存在很多技术和特定领域方面的问题，使得这个方向的研究非常具有挑战性.已有的聚类算法对数据对象的形状和密度有要求，应用到医学图像聚类方面不能取得很好的结果.针对以上问题，在领域知识指导下，首先对图像进行了纹理检测，提出了面向纹理的T-LBP方法;然后对预处理之后的图像进行了空间划分，并对每个空间内的纹理求取LBP值，建立按空间序列排序的LBP直方图;最后，将以LBP直方图作为特征，提出了基于图模型的医学图像聚类算法.实验结果表明，该算法在时间复杂度和聚类结果方面具有良好的效果.

Medical Image Clustering Algorithm Based on Graph Model

Clustering algorithm of medical image is a significant part of special field image clustering. Due to technical limit and many problems in specific area, the study in this direction has been very challenging. The exiting algorithms of clustering require shape and density of data object, which imply that there won't be a good outcome for the application of medical image clustering. To solve the problem above, this paper firstly detects texture from image, proposes T-LBP method, divides the preprocessed image into multiple spaces, calculates the value of LBP spaces, and then builds a spatial sequence LBP histogram. In the end, the clustering method of MCST is proposed based on the created LBP histogram. The outcome of this experiment indicates that the algorithm presented in this paper achieved good results in terms of time complexity and clustering function.