基于K-Means++聚类的体绘制高维传递函数设计方法

摘要

如何将体数据中重要的信息高质量地绘制出来是医学可视化急需解决的问题.基于高维直方图的高维传递函数交互设计法是目前流行的方法,但是该方法设计复杂且效果不理想.针对高维特征的传递函数设计问题,提出一个基于改进的K均值(K-Means++)聚类的高维传递函数自动设计与交互式的体绘制方法:首先,对三维数据场进行特征提取;然后,采用基于K-Means++聚类的传递函数自动生成方法;最后,提供便捷的交互式界面给用户进行调整.还利用基于图形处理器(GPU)的体绘制方法,充分利用图形卡的强大并行计算能力,达到实时绘制的效果.实验结果表明,该方法能消除高维传递函数设计的复杂性,并且能有效地融合多种人体组织结构特征,提高渲染效果.%The problem how to render the important information of the volume data qualitatively in medical visualization needs to be resolved urgently. The method of designing high dimensional transfer function interactively according to the high dimension histogram was used widely, but this method is complex and is of low quality. To solve the design problems of the characteristic high dimensional transfer function, a volume rendering method of automatic and interactive design of high dimensional transfer function based on X-Means ++ clustering algorithm was presented in this paper. Firstly, feature extraction was done on the volume data, then to cluster the feature space, X-Means ++ clustering algorithm was used, and the group of label transfer function was generated automatically. Finally, a convenient interactive user interface was provided to users to adjust. The GPU (Graphic Processing Unit) based volume rendering was used to perform the strong parallel computing ability for real-time rendering. The experimental results show that this method can eliminate the complexity of the design of high dimensional transfer function, and many kinds of human body organization structure characteristics can be shown in the rendering results.