Memory Efficient and Robust Software Implementation of the Raycast Algorithm

Abstract

In this paper we propose two novel software implementations of the ray-casting volume rendering algorithm for irregular grids, called ME-Raycast (Memory Efficient Ray-casting) and EME-Raycast (Enhanced Memory Efficient Ray-Casting). Our algorithms improve previous work by Bunyk et al [1] in terms of complete handling of degenerate cases, memory consumption, and type of cell allowed in the grid (tetrahedral and/or hexahedral). The use of a more compact and non-redundant data structure, allowed us to achieve higher memory efficiency. Our results show consistent and significant gains in the memory usage of ME-Raycast and EME-Raycast when compared to Bunyk et al implementation. Furthermore, our results also show that handling of degenerate cases generates accurate images, correctly rendering all the pixels in the image, while Bunyk et al implementation fails in rendering up to 38 pixels in the final image. When we compare our algorithms to other robust rendering algorithm, like ZSweep [2], we have considerable performance gains and competitive memory consumption. We conclude that ME-Raycast and EME-Raycast are efficient methods for ray-casting that allow in-core rendering of large datasets with no image errors.