3D RECONSTRUCTION AND APPLICATIONS

F. Galpin,

contact: F. Galpin, L. Morin

Goal

The goal of this work is the video sequence compression for very low bitrate. We assume static scene captured with a monocular moving camera. We propose a method using several 3D models in order to reconstruct the original video sequence or a virtual one. A complete scheme for video sequence compression is presented and compared with classical compression schemes.

Approach

The main idea is that instead of aiming at reconstructing a unique realistic model of the real 3D scene which needs too much assumptions, the video sequence is processed to produce several 3D models which are valid for a section of the sequence called GOP. These 3D models are then used for reconstructing the original video sequence with a very low bitrate.

See also previous results on reconstructed models and reconstructed videos and a global scheme of the method .

Discussion

Street sequence is provided by Thomson/CCETT.

Figure above shows the PSNR for both coding schemes.

Figure above shows a detail of the street sequence. Even with lower PSNR, the Rec3D method shows better visual quality : H263 introduces blocking effect and low-pass filtering whereas Rec3D method preserves details (car). On the other hand the Rec3D method introduces geometric distortions (on the left side wall).

Experimental Results

Examples of video sequence reconstruction

You can get a VRML plug-ins for netscape at the following URL

You can get a Unix VRML plug-ins for netscape at the following URL

You can get a Unix MPEG2 plug-ins for netscape at the following URL

References

[Galpin00d]

F. Galpin, Luce Morin "Computed 3D Models for Very Low Bitrate Video Coding" VCIP, 2000 (downloadable: PS version)