An Information‐Theoretic Observation Channel for Volume Visualization

Abstract

Different quality metrics have been proposed in the literature to evaluate how well a visualization represents the underlying data. In this paper, we present a new information-theoretic framework that quantifies the information transfer between the source data set and the rendered image. This approach is based on the definition of an observation channel whose input and output are given by the intensity values of the volumetric data set and the pixel colors, respectively. From this channel, the mutual information, a measure of information transfer or correlation between the input and the output, is used as a metric to evaluate the visualization quality. The usefulness of the proposed observation channel is illustrated with three fundamental visualization applications: selection of informative viewpoints, transfer function design, and light positioning.

What May Visualization Processes Optimize?

“…This holistic nature of information-theoretic reasoning has enabled many applications in visualization, including light source placement by Gumhold [33], view selection in mesh rendering by Vázquez et al [64] and Feixas et al [22], view selection in volume rendering by Bordoloi and Shen [5], and Takahashi and Takeshima [56], focus of attention in volume rendering by Viola et al [66], multi-resolution volume visualization by Wang and Shen [68], feature highlighting in unsteady multi-field visualization by Jänicke and Scheuermann [35,37], feature highlighting in time-varying volume visualization by Wang et al [70], transfer function design by Bruckner and Möller [10], and by Ruiz et al [8,49], multimodal data fusion by Bramon et al [6], evaluating isosurfaces [74], measuring of observation capacity [7], measuring information content in multivariate data [23], and confirming the mathematical feasibility of visual multiplexing [15].…”

An Information-Theoretic Framework for Evaluating Edge Bundling Visualization

“…Figure 2 illustrates , and , and their relationship. Many visualization studies [ 58 , 59 , 60 , 71 , 72 , 73 , 74 , 75 , 76 , 77 ] have proposed solutions to maximize the mutual information to improve their visualization results.…”

Rate-Distortion Theory for Clustering in the Perceptual Space

“…By optimizing the information transfer of this channel, some automatic solutions arise to design problems related to volume rendering, such as view-point selection and transfer function definition. It can be seen that the information channel proposed in [13] roughly corresponds to the vis-encoder subsystem proposed in [11] for the volume rendering technique. In our case, we will also consider only the vis-encoder subsystem, due to the availability to quantify both the input and output variables involved in this information channel.…”

“…From this approach, the visualization process can be studied from the perspective of information transfer or as an "information discovery process". A similar perspective was presented in [13], where the authors define an information channel between the original data and the colors depicted by the volume rendering visualization. By optimizing the information transfer of this channel, some automatic solutions arise to design problems related to volume rendering, such as view-point selection and transfer function definition.…”