Analyse et synthèse temps réel du visage pour la télévirtualité


Image(s): 640*480

Jpeg Image (32 Ko) Jpeg Image (20 Ko)
Jpeg Image (30 Ko) Jpeg Image (10 Ko)



Project : Télévirtualité

  • Email :
  • URL :

    Video(s) and extracted images: 320*240

    " Literature"
    Video QuickTime -> Film/Video (3.7 Mo)
    Jpeg Images -> (7 Ko) (6 Ko)

    " The mouth"
    Video QuickTime -> Film/Video (4.7 Mo)
    Jpeg Images -> (10 Ko) (10 Ko) (10 Ko)

    " The head"
    Video QuickTime -> Film/Video (7.0 Mo)
    Jpeg Images -> (10 Ko) (10 Ko)

    " The eyes"
    Video QuickTime -> Film/Video (3.5 Mo)
    Jpeg Images -> (11 Ko) (11 Ko) (8 Ko)

    " Expressions"
    Video QuickTime -> Film/Video (4.0 Mo)
    Jpeg Images -> (11 Ko) (11 Ko)

    " Laughing"
    Video QuickTime -> Film/Video (1.6 Mo)
    Jpeg Images -> (10 Ko) (11 Ko)


    A synthetic face is animated in real time from a real face. The spectator can see both faces simultaneously.

    Technical Information

    This film has been realized on a SGI indigo for the image capture and analysis and with an extrème SGI for the synthetic face rendering.

    More Information...

    • Bibliography :

      (2) "Real-time facial analysis and synthesis for televirtualite", Imagina's Proceedings pp173, 1994

    • Abstract :

      Although the animation of virtual clones is no longer just a dream, current drawbacks limit possible applications (contact device, sensor, head immobilization in front of the camera, etc.). In this article, we propose a real-time facial animation analysis/synthesis chain which imposes few constraints. This study has been elaborated for a televirtuality context: several physically distant persons intervene in a common virtual environment via their synthetic representatives. They are thus able to share a virtual community, enter a virtual minitel, follow a visioconference, access telework operations, or even join in multi participant games. The system can be broken down as follows:
      1. Video input
      2. Real image analysis: extraction of characteristic parameters
      3. Symbolic interpretation of parameters
      4. Transfer of these parameters to a remote machine
      5. Animation of the symbolic or realistic computer graphics clone
      6. Video output
      The video image processing is designed to extract real facial control parameters. This chiefly consists of identifying head movements and mouth deformations. For the realistic model, parameters corresponding to facial muscle movements derived from the analysis are interpreted in terms of muscular intensity. Only the parameters necessary for synthesis of the clone image are transmitted via the network or telephone line. The computer graphics clone may be either symbolic, in which case it is based on a geometric model, or realistic, in which case it is endowed with a muscular basis. In the latter case, the facial surface is represented by a network of springs. Each muscle is modeled by forces, and muscular deformations are calculated by solving dynamic equations for the given system (face/force). The muscular deformations are then pre calculated to obtain a system capable of real-time evolution. Finally, facial texture is mapped onto the polygonal representation of the face.

    • Some external links :

      (oo) Televirtutality project
      (oo) Facial Animation
      (oo) Facial Analysis

    • Some internal links :

      (oo) Same Author
      (ooo) Facial animation and ageing

  • Copyright © 1994-2024
    Other Sites : | Ai Girls | Ai Creations