Evaluating the Influence of Stereoscopy on Cluster Perception in Scatterplots

Abstract

Unlike 2D scatterplots, which only visualize 2D data, 3D scatterplots have the advantage of showing an additional dimension of data. However, cluster analysis can be difficult for the viewer since it is challenging to perceive depth in 3D scatterplots. In addition, 3D scatterplots suffer from overdraw and require more time for perception than their 2D equivalents. As an approach to this issue, stereoscopic rendering of three-dimensional point-based scatterplots is evaluated through a user study. In detail, participants’ ability to make precise judgements about the positions of clusters was explored. 2D scatterplots were compared to non-stereoscopic 3D and stereoscopic 3D scatterplots. The results showed that performance in perception decreased when confronted with 3D scatterplots in general, as opposed to 2D scatterplots. A tendency towards an improvement of perception showed when comparing stereoscopic 3D scatterplots to non-stereoscopic 3D scatterplots.