[Summaries Week 7]: Scape, Multi-viewer tabletop auto-stereoscopic display

Scape: Supporting Stereoscopic Collaboration in Augmented and Projective Environments

In the paper, Hua et al. talk about the design of a system that allows for collaboration in augmented environments. Their proposed system, Scape, allows for users to engage in face-to-face cooperation, while also providing each user with her unique stereo perspective for the augmented scene.

The authors mention how their motivation for designing the system stemmed from the inability of the state of the art to provide the desired experience. CAVE like systems permitted collaboration, but the rendered projections are driven by a single user’s perspective. Interleaving stereo perspective views for a couple of users led to a corresponding decrease in the display frame-rate, while the IllusionHole had a strict upper bound on the maximum number of people, post which there was bound to be cross-talk due to overlapping viewing frustums.

Scape tackles these issues by the use of Head Mounted Projection Displays(HMPDs) and retro-reflective surfaces. Using workspaces fitted with auto-reflective surfaces, the authors were able to provide outside-in views perspectives. Additionally, leveraging HPMD technology and strategically placed auto-reflective surfaces, the authors were also able to realize CAVE-like spaces that provided inside-out immersive perspectives.

While the use of auto-reflective material does place a bound of the area of the room where the experience is realized, the authors and the screen shape, this also provides intrinsic occlusion cues. A drawback to the technology, however, is that the auto-reflective materials need to be strategically placed in accordance with the experience desired.

The authors describe their hardware that makes Scape possible – workbench, retro-reflective immersive cage, multiple tracked HMPDs and multimodality interface devices. They also employ a vision base tracker to support interaction, and also outline an interface device – the magnifier, that can serve to magnify a particular section of a viewpoint. The authors also describe their software implementation that keeps track of states for various users, modeled as Actors, allowing them to enforce privacy policies on a per user basis.

Finally, the authors describe an example of their system in action – the Aztec Explorer that allows users to overview and discuss the city plan as well as experience the CAVE like walkthrough.



A Practical Multi-viewer Tabletop Auto-stereoscopic Display

In their paper, Ye et al. describe their multi-viewer tabletop auto-stereoscopic display that builds on Random Hole Displays (RHD), handling “colliding” pixels by proving for a real time GPU implementation of color blending and approximate error-diffusion.

The authors first describe the RHD and how its Poison disk distributed pseudo-random hole pattern handles cross-talk errors by error-diffusion due to its non-uniform hole pattern. They also point out the open issues in the technology that they seek to address, namely, no support for sub-pixel coverage handling, lack of user tracking and inter-reflection between barrier and display.

Highly accurate tracking (to the order of 5-10 mm) allows the rendering subsystem to handling sub-pixel level granularity for determining the final color. Additionally, the authors propose a multi-view blending algorithm for handling pixel conflicts that obtains the resulting color by taking the average of pixel colors in the individual views. The authors also experimented with other decision strategies for blending multiple view pixels, including, random selection and priority based selection based on user priority.

The authors then describe their five-pass hardware accelerated algorithm and describe how they track the eye locations using two retro-reflective tracking markers positioned at the wearer’s temples. The tracking algorithm includes passes to create individual view frames, blend them together and also a pass to diffuse the error over neighborhood pixels in the same view.

At the end, the authors discuss the possible applications – interior design collaboration, etc. for their system. They also measure the color errors reported by their system, commenting, that, while not brilliant, their results are an improvement over current state of the art. The authors also evaluate the improvement in performance of their display as a result of error diffusion, while also quantifying the degradation caused by additional viewpoints.