Large datasets have started emerging in many research fields. Information is being gathered on different levels of granularity. The graph metaphor is among the commonly used methods to visualise relationship between data elements in collected datasets. Adding the different granularity information into visualisations produces complex visual elements and structures. Visualisations are not very useful unless coupled with navigation methods and interaction techniques, otherwise users are not able to make practical use of such visualisations especially when the visualisation is performed on several levels of detail.
This thesis investigates navigation methods of large datasets that are visualized on several levels of detail in an immersive 3D virtual environment. The Focus+Context concept established in literature for 2D navigation is first extended into 3D and then applied on the camera metaphor, which resembles the existence of the user inside an immersive environment. A navigation and interaction framework is developed for navigating visual graph structures in 3D. The framework is tailored to observe human vision and perception limitations while computing camera paths in 3D space.
Five visual indices are devised based on concepts and established fact in Perception Psychology, Cognition, HCI and Graph Drawing to measure image quality and quantify the qualitative aspects of camera paths. A mixed approach between near exhaustive experimentation and mathematical analysis is used to study the behaviour of these visual indices on camera paths. Depending on their behavioural patterns for three classes of graphs, namely random, scale-free and clustered graphs, the visual indices are then incorporated into the navigation and interaction framework to produce optimised camera paths.
The navigation and interaction framework is demonstrated for navigation tasks through task based scenarios using a real world dataset, namely a metabolic pathway network extracted from the KEGG database. Suitable combinations of paths and visual quality measurements are suggested towards the end of this thesis for navigating the three classes of graphs in 3D environments.