Toed-in vs Parallel Displays in Video See-Through Head-Mounted Displays for Close-Up View

Abstract

In non-orthostereoscopic video see-through (VST) head-mounted displays (HMDs), the perception of the three-dimensional space is negatively altered by geometrical aberrations, which may lead to perceptual errors, problems of hand-eye coordination, and discomfort for the user. Parallax-free VST HMDs have been proposed, yet their embodiments are generally difficult to create. The present study investigates the guidelines for the development of non-orthostereoscopic VST HMDs capable of providing perceptually coherent augmentations for close-up views, hence specifically devoted to guide high-precision manual tasks. Our underlying rationale is that, under VST view, a perspective-preserving conversion of the camera frames is sufficient to restore the natural perception of the relative depths around a pre-defined working distance in non-orthostereoscopic VST HMDs. This perspective conversion needs to account for the geometry of the visor and the working distance. A simulation platform was designed to compare the on-image displacements between the direct view of the world and the perspective-corrected VST view, considering three different geometrical arrangements of cameras and displays. A user study with a custom-made VST HMD was then conducted to evaluate quantitatively and qualitatively which of the three configurations was the most effective in mitigating the impact of the geometrical aberrations around the reference distance. The results of the simulations and of the user study both proved that, in non-orthostereoscopic VST HMDs, display convergence can be prevented, as the perspective conversion of the camera frames is sufficient to restore the correct stereoscopic perception by the user in the peripersonal space.