摘要
提出一种新的基于笔画的传输函数设计方法。与其他基于笔画的方法相比,该方法只要求用户对感兴趣数据做少量标记,操作更简便;另外该方法更好地考虑了体数据的空间信息,分类结果更合理。首先使用贪心极大似然估计算法构建用户感兴趣数据和背景数据的高斯混合模型(GMM)模型;然后基于GMM模型使用Graph-Cuts实现对体数据的分类,并通过仔细设计Graph-Cuts能量函数的平滑项将空间信息引入分类过程。针对体数据的规模大,难以直接应用Graph-Cuts算法的难题,在此通过使用层次模型减小Graph-Cuts算法所处理问题的规模,从而获得了实时交互的效果。该方法使用多个体数据进行测试,获得令人满意的效果。
A new design method of paint-based transfer function is presented in this paper. In comparison with other paint- based methods, the proposed method only asks the user to mark the interesting data. It is more convenient anyway. In addition, the method more appropriately considers volume data＇s spatial information, and results in more reasonable classification. First-ly, the Gaussian mixed models （GMMs） for both user＇s interesting data and background data are established by greedy maxi- mum likelihood estimation algorithm, then classification is achieved by Graph-Cuts optimizations based on the established GMMs, and spatial information is involved in the classification through careful design of smoothing term of the Graph-Cuts ener- gy function. In general, volume data is too large to apply Graph-Cuts algorithm directly for interactive rate, which is resolved by using volume data＇s hierarchical model. Testing on several volume data has demonstrated the effectiveness and advantage of the method.
出处
《现代电子技术》
2013年第14期4-9,共6页
Modern Electronics Technique
基金
国家自然科学基金:多层次姿态可变辐射虚拟人模型及其建模方法研究(11005028)
国家自然科学基金:面向样例创新的动画角色重构研究(61070124)
安徽省自然科学基金:面向非真实感体绘制的智能界面的关键技术研究(090412066)
