| dc.description.abstract | Visualizations that show the shape of and spatial relationships between layers of surfaces are useful to oceanographers studying water masses or oncologists planning radiation treatments. The shape of and distances between layers is effectively visualized by displaying the surfaces semi-transparently and sparsely covered with opaque markings. However, it becomes difficult to distinguish the shapes and differentiate between the markings when showing more than two surfaces. Further, finding optimal sizes and numbers of markings for the different layers, so as to best display the surfaces, requires tedious manual effort. This dissertation firstly investigates animation of the opaque markings as a means of enhancing these visualizations. Such a Kinetic Visualization approach has several potential benefits: the perceptual grouping effect of similar motion helps distinguish between markings on separate layers, occlusions are modulated as the markings move, allowing a viewer to assemble an integrated mental image of otherwise partially obscured surfaces, and markings that follow certain trajectories such as surface curvatures, contribute to a better understanding of shape. Markings are also spread out in relation to the view point, reducing complete occlusions.Secondly, a computational model of human perception of a single surface is extended to layered surfaces by modelling processes of perceptual grouping and surface completion, incorporating relatability criteria. This model is intended to mimic a person’s perception of layered surfaces, and is used to measure the effectiveness of our visualizations, within an optimization framework, allowing optimal visualization settings to be automatically determined. Visualization enhancements through animation were evaluated through a user experiment comparing pendulum-style rocking, static renderings and Kinetic Visualization on sets of two surfaces. This showed that rocking alone results in more accurate depth judgements, indicating that the “Kinetic Depth Effect” is not induced by Kinetic Visualization. A follow-up experiment revealed that a combination of rocking and Kinetic Visualization is more useful than rocking alone for feature identification tasks when displaying four layers. Our perceptual model was evaluated, in an experiment, in which sets of layered surfaces were displayed using a range of different visualization settings. Respondents recreated surfaces matching their perception. Comparing our model’s evaluation of the different visualizations showed a weak linear correlation to the accuracy of the participant’s perception of the surfaces. This research shows that modelling perception of layered surfaces is a grand challenge and highlights the foundational problem of predicting significant variation that may arise between non-homogenous participants. | en_ZA |
