We have developed a quick and easy to use image-based modeling system for the acquisition of both 3D geometry and visual appearance of real world objects. Currently, the camera pose is calibrated with a planar calibration pattern and the silhouettes of the object are used to obtain a coarse geometry. To account for errors in the recovered geometry and to allow photo-realistic rendering of the captured object we also extract an appearance model of the object from the images. This appearance model is a compact representation of "how an object looks" from a particular view direction, allowing for novel re-renderings of the object from arbitrary viewpoints (more details). These models can then be composed with other virtual objects and used in any application a traditional computer graphic model is used.
Three - tier model: Macro geometry, Meso displacements and Micro Dynamic textures
Adds detail to a regular triangulated macroscopic model using a displacement map for pixel-wise depth and a dynamic texture basis to modulate view dependent light and geometric sub-pixel residuals.
Jagersand, M., Birkbeck, N., Cobzas, D.
A Three-tier Hierarchical Model for Capturing and Rendering of 3D Geometry and Appearance from 2D Images
,
International Symposium on 3D Data Processing, Visualization and Transmission (3DPVT) 2008
Presentation: Powerpoint, PDF
Videos
Videos
Concise and detailed overview videos of the capture system, and a video of several objects merged into a scene.- short system overview video (5 Mb mpeg4 format), (Same video mpeg2 format 26MB)
- detailed system overview video (50 Mb)
- composed scene (5 Mb)
Related Links
- virtual heritage content This is an example use of the system for the creation of virtual heritage content describing a inuit seal hunt.
- capture system user-guide
- detailed system description A self-contained overview of the system.
Project Contributors (past and present)
| Martin Jagersand | Dana Cobzas | Keith Yerex |
| Neil Birkbeck | Cleo Espiritu | Cam Upright |
| Adam Rachmielowski |