Identifying nearly equally spaced isosurfaces for volumetric data sets

Abstract

Isosurfaces are an important visual representation of volumetric data sets and isosurface extraction and rendering remains one of the most popular methods for volume visualization. Previous works identify a small set of representative isosurfaces from a set of sample ones, providing a concise description of the underlying volume. However, these methods do not lend themselves to equally spaced isosurfaces, i.e., keeping the same distance between neighboring isosurfaces, which can be advantageous from the user’s perspective in terms of visual summarization and interactive exploration. In this paper, we present a new solution that efficiently identifies a set of nearly equally spaced isosurfaces for a given volume data set. Our approach includes an estimation stage of linear interpolation and a refinement stage of binary search in order to balance the tradeoff between quality and performance. The refinement stage can incorporate spike and/or jump treatments to possibly improve the convergence. Experimenting with multiple data sets of different sizes and characteristics, we perform both quantitative and qualitative studies, demonstrate the efficiency and effectiveness of our approach, and summarize our findings.