image-based techniques for the creation and display of photorealistic interactive virtual...
DESCRIPTION
previous approaches n Limited level of realism in VE’s –creation of environments slow, costly and with limited realism –display of environments with low realism n missing shadows and vegetation, flat textures n missing sound n no life (crowds, traffic etc.)TRANSCRIPT
Image-based Techniques for the Creation
and Display of Photorealistic Interactive Virtual Environments
George Drettakis1, Maria Roussou3,4, Nicolas Tsingos1, Alex Reche1,2, Emmanuel Gallo1,2
1REVES/INRIA, Sophia-Antipolis, France2CSTB, France
3University College London, UK4Foundation of the Hellenic World, Hellas (Greece)
overall goals Realistic virtual environments (VE’s)
– simplify model creation– enhance realism of display– interactivity
Education and design– enhance learning process– help design and decision making
previous approaches Limited level of realism in VE’s
– creation of environments slow, costly and with limited realism
– display of environments with low realism
missing shadows and vegetation, flat textures
missing sound no life (crowds, traffic etc.)
previous approaches Many developments in Computer
Graphics, but rarely transferred to VE’s– considered too expensive in resources– engineering aspect overwhelming
Examples of CG techniques– image-based modeling [Debevec 96/98]– point-based rendering [Stamminger 01]– perspective shadow maps [Stamminger 02]
overview of our approach Project context
– CREATE IST project Three components
– creating and displaying a realistic VE– interactivity– user-centered design and evaluation
data acquisition for CREATE
Two applications areas– cultural heritage and education– urban planning and design
Two corresponding sites– Ancient Messene in Greece
(archeology)– Place Massena & Garibaldi in Nice
(tramway project)
model creation Image-based modeling
– shoot panoramas, stitch and render cube maps
resulting model After calibration, modelling &
texturing
synthetic image
video Creation of model using Stitcher
and ImageModeler (improved versions)
view-dependent 3D models
One common model, multiple textures– use blending for view-dependent display
For complex scenes use layering– perform geometric sort to create image
layers– use a view-dependent texture per layer
view-dependent 3D models
Results of Enhanced VE’s
urban planning models
synthetic image
enhanced virtual environment
enhanced virtual environment
video Place Garibaldi simulation
– shadows– sound
interactivity Cultural heritage
– manipulation of architectural members
– reconstruction of temple Urban planning
– design and construct “urban features” such as benches or umbrellas
user-centered design and evaluation
User task analysis identified expert and novice users:– Cultural heritage (C/H)
experts: archeologists novice users: general public, children
– Urban planning (U/P) experts: architects novice users: general public, decision
makers
user task analysis Interviews with users, analysis of
existing task flow– augment the current tools– add realism and interactivity to
enhance productivity Preliminary studies
– cultural heritage with children– preparation for urban case
preliminary formative studies
Children in a 3-task experiment (C/H)– 7 to 12 years old– CAVE environment
Preparation of usability evaluation with architects (U/P)– direct observation– description of needs– scenario-based evaluation
current work Completion of enhanced VE tools
– more realistic trees– integration of crowds and sound
Real-task experiments– architects in a design task for urban
planning (under way). Will be evaluated for effectiveness, efficiency, ease of use, and satisfaction.
video Interactivity
– design and placement of benches and umbrellas
conclusions Image based techniques
– improve realism of VE’s improved capture & display
– combined with interactivity and user-centered design show promise for improved usage of VE’s
acknowledgements CREATE is a 3-year RTD project funded by the
EU’s FP5 IST Programme (IST-2001-34231), http://www.cs.ucl.ac.uk/create. The partners of CREATE are University College London (UK), Foundation of the Hellenic World (GR), INRIA (FR), REALVIZ (FR), University of Cyprus (CY), PERCRO (I) and CSTB (FR)
Alias for their donation of Maya Alexandre Olivier-Mangon for all the image-
modeling and virtual modeling, and the videos Alex Reche and Emmanuel Gallo are supported
by CIFRE doctoral fellowships INRIA-CSTB
thank youhttp://www-sop.inria.fr/reves
http://www.cstb.frhttp://www.cs.ucl.ac.uk/create