Image-based X-ray visualization techniques for spatial understanding in outdoor augmented reality

This paper evaluates different state-of-the-art approaches for implementing an X-ray view in Augmented Reality (AR). Our focus is on approaches supporting a better scene understanding and in particular a better sense of depth order between physical objects and digital objects. One of the main goals of this work is to provide effective X-ray visualization techniques that work in unprepared outdoor environments. In order to achieve this goal, we focus on methods that automatically extract depth cues from video images. The extracted depth cues are combined in ghosting maps that are used to assign each video image pixel a transparency value to control the overlay in the AR view. Within our study, we analyze three different types of ghosting maps, 1) alpha-blending which uses a uniform alpha value within the ghosting map, 2) edge-based ghosting which is based on edge extraction and 3) image-based ghosting which incorporates perceptual grouping, saliency information, edges and texture details. Our study results demonstrate that the latter technique helps the user to understand the subsurface location of virtual objects better than using alpha-blending or the edge-based ghosting.