Displaying Stereoscopic Images

Download Report

Transcript Displaying Stereoscopic Images

My Categorization
 Free-Viewing Displays




SIRDS
Stereo Pairs
Barrier-Strip
Lenticular
 Aided-Viewing Displays
 Anaglyph
 Polarized
 Field-Sequential
Tradeoffs Considered
Cost  How easy/cheap is it to construct?
 How easy is it to view?
Usability
 How pronounced is the effect?
Effectiveness
 How many people can view the
Multi-viewer display with stereopsis at the same
time?
Animation  How easy is it to make an animated
version of the display?
Displays for the naked eye
 Multi-viewer is easy because people come
naturally equipped
 Cost, usability, effectivness, and animation vary
greatly




SIRDS
Stereo Pairs
Barrier-Strip
Lenticular
Single Image Random Dot
Stereograms (SIRDS)
 Commonly known as “Magic Eye”
 Appear to be noise -- they are! (with
constrains)
 Guide dots (if provided) indicate propert
convergence depth
 Only depth cue is stereo-disparity so the
stereo-blind (10% of population) never see
anything but noise!
Remarks





Notoriously difficult to view
Encode little visual information
Depth data is quantized (integral pixel offsets)
Extremely cheap to produce (with a computer)
Animation is possible (makes them easier to view
as well)
Stereo Image Pairs
 Simplest form of autostereograms
 Landmarks in image act a guides to aid in
finding proper convergence
 More angular adjustment of eyes is required
than in SIRDS
 Higher image quality at the cost of more
difficult viewing
Remarks
 Simplest to produce (darkroom, hand,
software,etc.)
 Compelling depth effect
 Viewable by many people at once
 High-strain with extended viewing
 Strain limits animation
Barrier Strip Displays
 Making viewers consciously adjust their
ocular convergence is uncomfortable for
some, impossible for others.
 Barrier strip displays use a grill of
occluding elements to block view of images
from either eye
 Viewers must be in certain locations to see
effect (angle and distance are tuned)
Note that barrier spacing is different than image slit spacing
Remarks
 Encode clean stereo disparity information
 Comfortable for extended viewing (natural
convergence point)
 Barriers block 50% of light going in and out,
usually requres backlighting
 Harder to construct (ugly trig)
 Rigid and expensive (structure requred to maintain
barrier spacing)
 Animation is no harder than still
 Commercial equipment available for medical
imaging
Lenticular Displays
 Defeat brightness problem of BS by controlling
ray path with lenses instead of barriers
 Array of long cylindrical lenses (per pixel column)
refract light to places with same distance
constraint as BS, continuous angle
 100% of light passes in and out, no backlighting
necessary
 Wider field of view (limited by TIR and selfocclusion)
Remarks
 Animation is possible with still source
images using motion of viewer
 Able to ~reproduce lightfield
 More expensive/complex than BS with
higher quality and less contraints
 Drop-in graphics libraries can turn any 3d
program into a lenticular display source
Displays with special viewing
hardware
 Hardware can enable better {usability,
effectiveness, multi-viewer, animation} at the cost
of cost -- the normal technology vs nature
tradeoff.




Anaglyph
Polarized
Field-sequential
Dual display
Anaglyph
 Nerdy/Cool red-blue glasses
 Cyan, not blue!
 Two images overlap (like SIRDS) but are
differentiated by color
 Filters over each eye collect light from one
image but not the other
 Works based on intensity of light -colorblind people see them fine!
Remarks








Convergence is natural
Crosstalk can be annoying
“Color bombardment” causes strain and after-effects
Strain limits long term viewing
Same depth resolution/quality as raw stereo pair
Small incremental cost
Easy to make with (software/hand)
Animation is easy
Polarized Displays
 Approach is similar to anaglyph
 Polarization differentiates L-R channels
 Requires two polarized light projectors
(instead of just a printed page)
 Screen must be polarization-preserving
 Light loss and crosstalk occur when uses tilt
head
Remarks




The cost-wise step up from anaglyph
Completely natural viewing experience
No strain (unless glasses cramp your style)
Ideal for theaters (IMAX), because high upfront costs and low incremental costs
Field Sequential Displays
 Polarized projectors and screens do not make
economic sense on a single-user scale
 Move system complexity to the glasses from the
display
 LCD shutters over each eye control light flow
from conventional display (monitor/projector)
 Inexpensive control box triggers shutter
 Several (expensive) glasses can be driven by one
control box
Liquid Crystal Shutter Glasses
L
R
End of each scan-line.
Remarks
 Convergence is natural (still)
 Some crosstalk can occur with lingering
phosphors, slow shutters, synchronization
issues
 Cost is proportional to the number of
viewers
Dual Displays
 Enough monkey business, just stick a
monitor in front of each eye.
 Heavy (and expensive) headgear provides
bright, immersive experience
 Can be combined with headphones and
head tracking to modify experience based
on head movement
Nerd.
Remarks
 Expensive
 Completely natural focus (lenses embedded
in headgear)
 Very effective
 Animation is standard
 Only one user at a time
 Prices are dropping
Conclusions
 Noooooo! My awesome comparison matrix is
gone!
 Usability
 Lenticular and dual displays are best
 Effectiveness
 SIRDS and anaglyph are the worst
 Multi-viewer
 Barrier-strip and dual displays have the most constrains
 Animation
 Its always possible but strain limits application to
videos