【摘要】 为了从MRJ三维采样数据生成空间中任一位置及任一方向的切片图像,我们在物体空间中及计算机屏幕上建立了两套坐标系,引入了六个参数来描述切割平面.推导了从屏幕坐标到物体空间坐标的映射公式,设计了六种密度估计算法,即三线性插值法,最近邻法、中值法、控制力法,梯度法及GNP综合法,用于从所给的数据来估计空间中任意位置的密度,所有的算法都在某些情况下显现了它们的优点. 我们建立了—个由10个尺寸、方向,密度各不相同的椭球组成的三维头模型,通过在物体空间的均匀采样来生成数据集.使用了多组参数来检验模型和算法的成像能力. 在对算法结果进行了主、客观的比较之后,我们总结了这些算法的优,缺点.对于—般的应用,我们推荐梯度法与GNP综合法,在大多数情况下,这两种算法都能产生平滑且明显的边界. 算法的测试与比较使用了我们自己编制的一个基于Windows 95的程序.更多还原

【Abstract】 We designed a model with six alternative algorithms to produce the slice of the three-dimensional array. The slice plane can have any orientation and any location in space.Two coordinate systems are set up in the object space and on the computer screen, respectively. Six parameters are introduced to describe the slice plane, and the coordinate mapping from the screen to the object space is formulated. The six density-estimating algorithms, i.e., the Trilinear, the Nearest-Neighbor, the Median, the Power-Control, the Gradient and the GNP-Integrated algorithms, are designed to use the given data to estimate the density of any location in space. All the algorithms show their advantages in some conditions.Our 3-D head model consists of 10 ellipsoids of different size, orientation and density. The data sets are sampled in the object (head) space at some even intervals. We devised several conditions to test our model and algorithms. Based on the subjective and objective comparisons between different algorithms, we summarized their strengths and weaknesses. For commonly use, the Gradient algorithm and the GNP-Integrated algorithm are suggested. Both algorithms can produce slices of sharp and smooth edges in most cases.A window-based program was designed to test and compare the results of different algorithms.更多还原