Computer Science and
     Software Engineering

Computer Science and Software Engineering

PHD 01/08

AR Magic Lenses: Addressing the Challenge of Focus and Context in Augmented Reality

Julian Looser
Department of Computer Science
University of Canterbury

Abstract

In recent years, technical advances in the field of Augmented Reality (AR), coupled with the acceleration in computer and graphics processing power, have brought robust and affordable AR within the reach of the wider research community. While the technical issues of AR remain heavily researched, there is also a growing amount of work on user interface development and evaluation, heralding the convergence of traditional Human Computer Interaction (HCI) and AR. Magic Lenses are 2D interface components that provide alternative representations of objects seen through them. In this way, they can be used to provide Focus and Context in the interface, especially when visualising layered information. There are very few, if any, formal evaluations to guide the development of lens-based interfaces. This thesis describes the development and evaluation of Magic Lenses as a tool for AR interfaces. The work starts with a comprehensive survey of many Focus and Context techniques, which are classified based on the way they present views to the users { for example, a Magic Lens is a spatially separated multiple view technique. A formal evaluation of 2D Magic Lenses in a GIS scenario found that users strongly preferred the lens-based interaction technique to others, largely because it reduced the effort of interaction. Accuracy was high with the lenses, but a simple “global view” interface allowed significantly faster performance. This positive result motivated further work on Magic Lenses within AR, where the lens metaphor can reinforce the tangible interaction methods that link virtual and real content. To support rapid exploration of interaction alternatives with AR Magic Lenses, I describe the design and architecture of osgART, an AR development toolkit that is available to the research community as open-source software. Object selection and manipulation is a fundamental interaction requirement for all AR interfaces, and I establish an empirical foundation of performance in this task with a variety of AR interaction techniques, including Magic Lenses. Results show that performance with all techniques is successfully modelled by Fitts’ Law, and that Magic Lenses outperformed other techniques. Finally, I examine new interaction techniques based on Magic Lenses, particularly a Flexible Sheet Lens, which allows concurrent bimanual specification of multiple parameters within the visualisation.