Continuous Parallel Coordinates (CPC) are a visualization technique used to perform multivariate analysis of different scalar fields defined on the same domain. While classic Parallel Coordinates draws a line for each sample point, a CPC visualization uses a density-based representation. An interesting possibility for the classic method is to highlight higher-dimensional clusters using edge bundling, where each line becomes a spline bent towards the centroid of the cluster. This often leads to expressive, illustrative visualizations. Unfortunately, bundling lines is not possible for CPC, as they are not involved in this method. In this paper, we propose a deformation of the visualization space for Continuous Parallel Coordinates that leads to similar results as those obtained through classic edge bundling. We achieve this by performing a curved-profile transformation in image space. The approach lends itself to a computationally lightweight GPU implementation. Furthermore, we provide intuitive interactions with the bundled clusters. We show several examples of our technique applied to a commonly available data set.


List of all publications