: This paper presents a method to preserve sharp edge details in splatting for volume rendering. Conventional splatting algorithms produce fuzzy images for views close to the volume model. The lack of details in such views greatly hinders study and manipulation of data sets using virtual navigation. Our method applies a non-linear warping to the footprints of conventional splat and builds a table of footprints for different possible edge positions and edge strengths. When rendering, we pick a footprint from the table for each splat, based on the relative position of the voxel to the closest edge. Encouraging results have been achieved both for synthetic data and medical data.