Intelligent Focus+Context Volume Visualization

摘要

Although graphics processing unit (GPU) acceleration makes possible interactive volume rendering, successful volume visualization relies on the ability to quickly and correctly classify the volume into different materials or features. Among various classification techniques, one very attractive and effective method is employing machine learning to classify the whole volume according to some minimum user input through an interactive brushing interface, where users paint directly on slices of the volume. For routine visualization tasks, we can thus reduce their cost if the visualization system can learn the tasks and apply the captured knowledge in future tasks. This paper presents an intelligent, interactive visualization system that supports Focus+Context viewing of volume data. Features of interest should be the focal point of the visualization, and by applying appropriate rendering methods we are able to enhance these features and create more illustrative visualizations in a Focus+Context style. We show with a set of case studies that it is possible to use machine learning to not only help classify volume but also better present the classified results. This new capability makes visualization a more usable tool.