Optimizing Visual Conspicuity of Data in 3D Volume Visualization
John Dingliana, SCSS
12-1pm 29th Mar 2019
In this talk, we present an approach for optimizing key parameters used in the visualization of 3D volumetric data. Such data, in the form of a vector or scalar field sampled over a discretized 3D grid, is ubiquitous in scientific data analysis, but, due to its visual complexity can lead to problems of information overload. This complexity can be managed by enhancing a subset of the data most relevant to the visual analysis task at hand, whilst extraneous information is de-emphasized. However, defining which parts of the data to enhance or reduce is a non-trivial problem, the solutions to which are often very specific to the given visual analysis task at hand.
In order to address this problem, we propose an automated computational metric, called Visibility-Weighted Saliency (VWS), that quantifies the conspicuity of sample points in the data when rendered with respect to a set of modifiable visualization parameters, specifically the user’s viewpoint and the mapping of data to specific colour and opacity values. By iteratively adjusting these parameters and scoring the resultant visualizations based on our metric, we can arrive at a set of parameter values that satisfies user-specified needs for a given analytics task. The approach can be applied to aid interactive visual exploration and spatial analysis of static and time-varying volumetric datasets.
This research was carried out in collaboration with Shengzhou Luo as part of the SFI funded project, ARTIVVIS: Real-time Time-Varying Volume Visualization (13/IA/1895).