Talking Heads mag from Barwon Heads Victoria AustraliaApril 2015
3D Talking Heads : Image Based Modeling at …welch/media/pdf/3dth_TR98-017.pdf · 3D Talking Heads...
4
UNC is an Equal Opportunity/Affirmative Action Institution. TR98-017 January1998 Department of Computer Science CB #3175, Sitterson Hall UNC-Chapel Hill Chapel Hill, NC 27599-3175 U N I V E R S I TA T C A R O L S E P T E N T S I G I L L U M • • • LUX LIBERTAS 3D Talking Heads : Image Based Modeling at Interactive rates using Structured Light Projection Ramesh Raskar, Henry Fuchs, Greg Welch, Adam Lake, Matt Cutts
Transcript of 3D Talking Heads : Image Based Modeling at …welch/media/pdf/3dth_TR98-017.pdf · 3D Talking Heads...
UNC is an Equal Opportunity/Affirmative Action Institution.
TR98-017January1998
Department of Computer ScienceCB #3175, Sitterson HallUNC-Chapel HillChapel Hill, NC 27599-3175
UN
IVERS I T AT
CAROL
SE
PTENTS
IG I
LL
UM
•
••
LUX
LIBERTAS
3D Talking Heads : Image Based Modeling atInteractive rates using
Structured Light Projection
Ramesh Raskar, Henry Fuchs, Greg Welch, Adam Lake, Matt Cutts
3D Talking Heads : Image Based Modeling atInteractive rates using Structured Light Projection
Real Time Depth Extraction
Ramesh Raskar, Henry Fuchs, Greg Welch, Adam Lake, Matt Cutts
University of North Carolina at Chapel HillDept of Computer Science
Technical Report TR98-017, 1998
The 3D talking head project at UNC allows participants at remote sites to be seen in 3D rather than asflat video images in application such as teleconferencing.
The collaborator atremote site isscanned by acombination ofactive lightprojector (DLP:Digital lightprojector) and asyncronized videocamera. The videocamera is acombination of twocameras, geometryand texture camera,which share theiroptical axis. Theacquired 3Dinformation aboutthe scene isrepresented as a 3Dcomputer graphicsmodel at the host.
file:///mac-welch/Users/Greg/Documents/Submissions/technical%20reports/3dth/start.html (1 of 3) [8/18/2000 7:55:34 AM]
The underlying difficultcorrespondence problem fromstereopsis is often reduced byprojecting structured lightpatterns. We currently usebinary coded vertical bars. Acamera looks at a set ofpatterns to distinguish betweenall the projected vertical bars.To distinguish between 512projected columns 9 patternsare required.
The 3D world coordinates ofpoints on the surface of sceneobjects are computed bytriangulation of the projectedplane of individual column ofthe projector and rays incidenton each camera pixel. Themajor challenge is to calibratethe environment and to set upthe calculations in such a waythat triangulation forcomputing 3D worldcoordinates is efficient. Wehave reduced this to 7 accessesfor lookup plus 3 adds and 6multiplications at each pixel.This can also be parallelizedvery easily.
We hope to make the projected patterns imperceptible to human eye. Binary image of the pattern and itsinverse are projected in quick succession. This appears as a uniform grey level. A camera synchronizedto the projector however can detect the pattern by capturing only the first half duration.
(Insert here two images with inverse patterns projected and also the same scene with constant fieldillumination.)
We anticipate that real time imperceptible extraction and display of 3D environments with people willnot only enhance the sense of presence in teleconferencing but will also inspire a variery of newapplications among which we plan to pursue telemedicine and distributed collaborative design.
file:///mac-welch/Users/Greg/Documents/Submissions/technical%20reports/3dth/start.html (2 of 3) [8/18/2000 7:55:34 AM]
Members :Advisors : Henry Fuchs, Greg WelchRamesh Raskar, Adam Lake, Matt Cutts
Support : National Science Foundation Cooperative Agreement no. ASC-8920219: "Science andTechnology Center for Computer Graphics and Scientific Visualization"
file:///mac-welch/Users/Greg/Documents/Submissions/technical%20reports/3dth/start.html (3 of 3) [8/18/2000 7:55:34 AM]
