Week 7 Summaries
A Practical Multi-viewer Tabletop Autostereoscopic Display
The authors introduce a multi-user autostereoscopic tabletop display and its associated real-time rendering methods. They use the “Random Hole Display” design that modified the pattern of openings in a barrier mounted in front of a flat panel display from thin slits to a dense pattern of tiny, pseudo-randomly placed holes. This allows viewers anywhere in front of the display to see a different subset of the display’s native pixels through the random-hole screen. Their main challenge is handling “conflicting pixels”. Autostereoscopic displays have been around for a while but they have not become popular due to their high cost to performance ratio. Their solution relies on head-tracking, supports 4 viewers at max, handles sub-pixel coverage issue and inter-reflection that causes significant interference. To achieve all this they have developed a hardware-accelerated rendering solution and novel calibration method tested on some sample applications like the room-designing application.
The barrier pattern was a randomly generated distribution of transparent boxes that were not aligned with the pixels. LCD pixel consists of RGB regions which will have these random holes on top of them. They used projective geometry calculations to account for refraction to determine which parts are visible. This was useful in solving the “sub-pixel” problem caused because holes were not aligned with the pixels. To do this, they computed which fraction of pixel are visible to the user but required extremely precise tracking of the eye. Because the images were down sampled, it required some pre filtering to reduce aliasing. To blend multiple views, they just averaged the colors in the visible areas. A 4-pass rendering algorithm was developed to render. It rendered the downsized image in the frame buffer, then generated the visibility mask for each user, then the 2 were combined and pixel values were computed and error map generated and finally the error is diffused to the nearby pixels. They suggested a probabilistic optimization which could resolve the conflicting pixels based on viewer’s priority. They needed extremely accurate position of the user’s eye-balls by using 2 retro-reflective spheres, which determine the alignment of your eyes and the reflecting-balls.
They developed several apps to test this like 3D reconstruction of a city, cubes and a room-design application. In general, the results were good, but they reduced as more users were added due to crosstalk (leading to significant noise). They supported graceful image degradation and enhancement depending on whether users were added or removed from viewing.
Scape: Supporting Stereoscopic Collaboration in Augmented and Projective Environments
The authors want to propose a collaborative tool for a bunch of people working together. Their main goals were for each to view individual perspective, natural access, perceive the presence of others and dynamically switch focus between shared-work and interpersonal space.
They intend to use Head Mounted Projective displays (HMPD) for which they developed their own prototype. It uses projective lenses, a beam splitter, a retro-reflective screen and a LCD display. The HMPDs help enhance the real world with 3D graphics, create multiple view points and at the same time allowing face-to-face communications. The HMPD only work on surfaces that are retro reflectors which limits the mobility of the system. They also made a CAVE-like system using the same reflective surfaces. Retro-reflective surfaces only tolerate small angles so they applied it to every wall and corner. The retro-reflective surfaces restrict the mobility but are awesome because they can be placed anywhere without causing blurring or degrading the images.
For interaction and tracking they used and tested several different options. They made work-benches for this. For example, in the case of interaction, you could use a glove or a standard mouse and keyboard. They also used tracking to create a Touring application in which one user was teleported into the view of another user with the walls showing view. They also made a cylinder app in which the views are different from different angles.
All in all, the system looks interesting and emphasizes on collaboration mentioned at the start seems to carry through the entire paper and work.
Comments are closed.