P3D 3D impressions from SID 98 (2 of 3)

[roberts@xxxxxxxxxxxxxxxxx (John W Roberts)]

[Part 2 of 3: my opinions]

S-22.3: "Tilting Disc: A Real Scale Interface for Cyberspace"
 - The speaker described a system being developed in his laboratory for
   a VR "navigation" system in which the user wears a stereo head mounted
   display, and moves through a virtual space by standing on a disk which
   is actually tilted by the motion of the user, the direction of "motion"
   matching the direction of the tilt. The user interacts with objects
   in the virtual environment, and with other users, represented as icons
   "avatars". The navigation system is patterned after Japanese (legends?)
   of a character who could move through the air.

S-32.1: "A Display System with 2D/3D Compatibility" (Sanyo)
 - The system described is an autostereoscopic display (no glasses
   required), built into an LCD panel and utilizing a barrier-strip
   type technology. There are both single- and double image splitter
   designs. In the single image splitter design, the splitter (alternating
   vertical apertures and barriers) is placed between the backlight and
   the LC panel, configured so that when the user is in the correct
   position, the light from the even-numbered pixels in each row reaches
   only one eye, and the light from the odd-numbered pixels reaches only
   the other eye. (The user moves his/her head around until the proper
   spot for viewing is located.) The barrier strips are reflective on
   the back, and there of course the usual mirror behind the backlight,
   so that much of the light that would be lost in a traditional barrier
   strip system is recovered. A 2D/3D compatible display can be built by
   placing a special shutter between the image splitter and the LC panel:
   the shutter can be electrically switched between two states - transmitting
   (transparent) and scattering (diffusing). In the scattering mode, the
   alignment of the light beams set up by the image splitter is disrupted,
   and the display acts as a regular 2D display. There is also a double
   image splitter design, which has one image splitter between the backlight
   and the LC panel, and one between the LC panel and the user - this reduces
   crosstalk between the left and right images, and improves 3D performance.
 - Sanyo's latest design, a 15" 1024x768 panel with double image splitters,
   was demonstrated at the author interviews. I thought it looked really
   spectacular - images shown included surgical photographs (surgery is one
   of the intended applications), and some phenomenal, highly detailed
   paintings by a Japanese artist (whose name I neglected to record).

S-32.2: "A Real-Time 2D to 3D Conversion Technique Using Computed Image 
   Depth" (Sanyo)
 - This arose as part of the 3D display program - due to the perceived
   shortage of 3D material, it was decided to try to develop an algorithm
   to convert existing 2D material to stereo in real time, and put it on
   a VLSI chip.
 - The chip utilizes two methods separately or in combination, depending
   on the cues that are available in the image:
    - Modified Time Difference (MTD): picks up image cues from lateral
      motion.
    - Computed Image Depth (CID): works on the concept that nearer objects
      generally have greater contrast and sharpness than more distant
      objects. The scene is divided up into a large number of squares,
      and an estimate made for the depth of each. The elements are then
      shifted differently in the left and right views, the amount of shift
      depending on the estimated distance (for example, a distant object
      would be shifted more to the left in the left view, and more to the
      right in the right view).
 - The computed left and right views are shown sequentially and viewed using
   LCS glasses. The system was demonstrated on a video including video and
   still images. I was pleasantly surprised at how well it worked. The
   perception of depth was not completely continuous and didn't work ideally
   for all still images, but most of the video shots and many of the still
   images showed depth in about the same way I would have judged it looking
   at a 2D display (except that it was the chip making the judgements).
   Response time to set the depth for a new scene or a rapid change in
   scene seemed to be about a second - this may have been partly a deliberate
   choice to avoid rapid changes in depth settings. The process seemed to
   work best with detailed scenes and good focus. Some classical 2D paintings
   converted amazingly well (which perhaps shows a compatibility between
   the painters' ideas of the representation of depth and Sanyo's.)
 - The speaker indicated that further improvements can be made, in handling
   replacement of the portions of the background that are lost by motion of
   objects in the foreground.
 - Automated 2D -> 3D conversion has been so universally condemned on P3D
   that I didn't have very high expectations, but I figured that most
   people claiming to achieve this haven't also invented their own
   autostereo LCD displays, so I ought to go see their demo. It worked
   considerably better than I had expected. The fact that there were scenes
   where it didn't work as well was not surprising - people would also have
   more trouble judging depth in some scenes than in others.

[Continued]


------------------------------

End of PHOTO-3D Digest 2746
***************************