Model-based visualization: Computing perceptually optimal visualizations

Abstract

Visualization is often more an art than a science. Here a model-based approach to visualization is promoted in order to make steps forward in this respect. The main ingredients of this approach are the derivation of a quantitative model of the perception of (certain aspects of) a visualization, the definition of an optimal visualization, followed by its computation. Some examples of this approach are presented, with optimal zooming and panning as main example. Advantages and problems are discussed. In conclusion, we think this approach to be challenging and promising, although we acknowledge that it is not an easy route.