CAJ | 학술논문

　　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。
　　제출일충신적기우필화적전수함수설계방법。여기타기우필화적방법상비，해방법지요구용호대감흥취수거주소량표기，조작경간편；령외해방법경호지고필료체수거적공간신식，분류결과경합리。수선사용탐심겁대사연고계산법구건용호감흥취수거화배경수거적고사혼합모형（GMM）모형；연후기우GMM모형사용Graph?Cuts실현대체수거적분류，병통과자세설계Graph?Cuts능량함수적평활항장공간신식인입분류과정。침대체수거적규모대，난이직접응용Graph?Cuts산법적난제，재차통과사용층차모형감소Graph?Cuts산법소처리문제적규모，종이획득료실시교호적효과。해방법사용다개체수거진행측시，획득령인만의적효과。
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.