A comparison of 2D and 3D interfaces for editing surfaces reconstructed from contours

Abstract

The last decade of computer technology has seen the proliferation of computer graphics applications. As technology advances, there is a growing fascination with three-dimensional (3D) object representations that likely comes from their greater ability to match "real life" than their two-dimensional (2D) counterparts. Unfortunately, the benefits of 3D editing are not without a price. Most techniques for manipulating objects in a 3D environment are developed for conventional hardware configurations that use 2D input devices and CRT displays. The difficulties lie in mapping 3D spatial relationships to 2D displays, an in mapping 2D user input to 3D object manipulation. This mapping problem is somewhat mitigated by adding constraints to the degrees of freedom in the manipulation task. 3D surfaces that have been reconstructed from contours are interesting to consider as targets of 3D interaction because they provide an inherent constraint on manipulation: point motion is restricted to a plane.

As part of my research, I implemented an interactive contour editor to edit 3D surfaces that were reconstructed from planar contours. More precisely, the editor is a tool for visualising a surface derived from a set of serial sections, and for removing deformations from this surface. It was designed specifically to remove artefacts from medical images of arteries.

I used the interface from my editor in an experiment that tested whether users were faster and more accurate at manipulating surfaces in a 2D environment or a 3D environment. At the outset of this study, I predicted that 2D would be better for editing deformations of a 2D nature. That prediction was borne out by my experimental results. I had also hoped that 3D would be superior as an editing environment for correcting deformations of a 3D nature. However, the 2D character of the data had a stronger effect on performance than did the 3D character of the deformation. Despite the inherent constraints in their surfaces, participants were faster at editing in 2D for all types of deformations, while maintaining a consistent accuracy between 2D and 3D. Participants did perceive a 3D environment to be better than a 2D environment for manipulating a group of points that spanned multiple contours, although this was not reflected in the quantitative results. The intuitive preference for 3D in this situation leads me to believe that it is worth continuing the search for a natural and effective interface for editing surfaces in a 3D environment.