面向体数据细节特征的传输函数设计方法研究

项目名称： 面向体数据细节特征的传输函数设计方法研究

项目编号： No.61303133

项目类型： 青年科学基金项目

立项/批准年度： 2014

项目学科： 自动化技术、计算机技术

项目作者： 周志光

作者单位： 浙江财经学院

项目金额： 22万元

中文摘要： 体数据细节特征的准确分析与增强展示，是辅助领域专家对体数据做进一步分析与解释的关键。然而，基于体数据细节特征尺度细微、体素数量少、属性分布不均、易受噪声干扰等特点，当前传输函数难以对其实现准确的分析与光学参数设计。本项目拟设计迭代式多维传输函数，综合考虑体数据多维属性信息，动态推荐最佳分类属性，迭代式地探索与分析体数据细节特征。在传输函数特征空间分类的基础上，引入机器学习方法，结合体数据空间位置及多维属性信息，对体数据细节特征进行空间划分，并且利用树状图有效表达特征之间的层次关系，实现体数据细节特征的完整分析与准确表达。进而设计局部像素级可见性分析方法，相对准确地描述体数据细节特征的视觉感知，实现传输函数参数的自动优化设计，有助于体数据细节特征的定量分析与增强可视化。将提供便捷的传输函数交互模式，实现集成体数据细节特征分析、提取及增强的体数据可视化原型系统，验证所提算法的有效性与实用性。

中文关键词： 体数据；细节特征；传输函数；机器学习；可见性

英文摘要： It is much important to extract and highlight the volume tiny features, that would be helpful for domain experts to do further analysis and interpretation for the volume. However, with current transfer functions, it is hard to conduct the effective classification and optical property design for the volume tiny features, because of their inherent characteristics, such as small scale, little size，uneven distribution of attributes and sensitivity to noise. In this project, a novel iterative multi-dimensional transfer function is designed for the accurately classification of the tiny features of interest, with a large number of valuable attributes considered. In order to further refine the feature classification based on transfer functions, machine learning is applied with the spatial information considered. The tree diagram is used to depict the hierarchical relationship for the features from the transfer function space and data space. With the above analysis of tiny features of interest, we further design local visibility analysis method to accurately achieve the visual perception of the tiny features. Accordingly, the optical property can be adaptively designed to enhance the visual perception of tiny features of interest. Finally, we provide a user-friendly interface of transfer function design, and develop a pr

英文关键词： Volume visualization；Tiny feature；Transfer function；Machine learning；Visibility