Generating 3D Mosaics from Stereo Image Sequences

Mircea Nicolescu


This work focuses on generating realistic 3D mosaics of a static environment from sequences of stereo image pairs, even without carefully calibrated or expensive camera systems. By reconstructing individual 3D representations for each frame as explicit surfaces and transforming them into a unique virtual viewpoint coordinate system, we are able to merge them into a 3D mosaic of the scene. A 3D Z-buffer algorithm may be used to integrate all surfaces into the final model and to discard points that appear in more than one frame. This method allows placing computer generated and real objects within the same 3D model for applications such as navigation in virtual environments. It can also be used in a virtual collaboration system, for automatic background generation.

On-line references

Generating 3D Mosaics from Stereo Image Sequences

Maintained by Alexandre R.J. FRANÇOIS