Curvilinear Grid Filtering by Adaptive Evaluation

Data from computational fluid dynamics (CFD) are frequently computed on curvilinear grids. In contrast to cartesian or uniform grids, the cell size in a curvilinear grid can vary significantly, which in turn complicates filter operations tremendously. Some of these filter operations may for example re-map the value range of the data values into another range. In this paper, we present adaptive filter evaluation that avoids the cartesian resampling of the dataset to take into account the underlying grid. Instead, we filter the voxels at the grid positions and correct the respective contribution to the convolution. As an example, we used and adapted VHDR, a high-dynamic range mapping operator that we previously used for cartesian grids.