Multi-View Surface Reconstructions of Objects [ ICCV'03, PAMI'05 ]

Email: Maxime { dot } Lhuillier { at } free { dot } fr
Address: Institut Pascal, UMR 6602 CNRS/UBP/IFMA,
Campus Universitaire des Cézeaux,
63171 Aubière Cedex - France.
Tel: +33 (0)4 73 40 75 93
Fax: +33 (0)4 73 40 72 62

Last modification on October 9th 2003.
Go back to my HOME PAGE.

About it

The algorithm input is an image sequence taken with an usual and hand held still camera, and the algorithm output is a set of textured triangles which approximate the object surface.

Very sketchly, an intermediate result is a cloud of 3D reconstructed points and the camera motion obtained by the method described in [ ECCV'02 ].
The next step is the progressive deformation of surface which approximates the set of 3D points, using sometimes 2D data in images.

Examples

Clic on the following icons to visualize a mpeg video. You can also download the resulting Vrml models and their textures. Each camera (resp. reconstructed 3D point) is represented by a black and white pyramid (resp. a textured square in the neighborhood of the point).

Bust Sequence: (26 images at 480x640)

Here are the Vrml model and its texture.

Ajay Sequence: (38 images at 480x640)

Here are the Vrml model and its texture
Thanks to Dr. Ajay for providing the input images !