The Solution of Data Gradient on Nonuniform Grid in Direct Volume Rendering
Jia Zhong, Ziming Zou · 2017
It is an important part of extracting the border in volume visualization for computing gradient feature of spatial data accurately and rapidly. The traditional method of volume visualization is to resample the data of model or observation distributing on non-uniform grids of three-dimensional Euclidean space by trilinear interpolation to uniform grids, then calculate gradient features of all grid vertex data. Based on the law of the change of partial derivatives along axis, this paper proposes a new method, to solve the gradient by mapping the partial derivative to a uniform grid space directly after solving it on a non-uniform grid. When the data value in three-dimensional space corresponds to quadratic function distribution, the new method, theoretically, has a higher accuracy than the old method. Finally, this paper verifies the algorithm validity with visualization experiment using the MHD simulation data to extract the magnetopause. Therefore, this paper argues that whether the law of the change of partial derivatives is linear or not and the step of grid along axis are the judgements of sequence of interpolation and feature calculation.