Integration of image-based objects in virtual environments

Version française

Official: Member: Team of XLIM laboratory:  
    Work initiated with Alain Fournier
    University of British Columbia, Imager Lab

Context

In this project, our goal is to insert image-based objects into virtual environments, using image-based rendering techniques combined with usual rendering techniqus. The main problems concern: object depth, object placement, surface reshading, shadows and so on.
First, we aim at recovering as much information as possible from a set of photographs (object shape, BRDF, light sources, etc.). Then, recovered information is combined with the original images for the object relighting and integration into virtual scenes.

System overview

Work

       
  • A Framework for Automatically Recovering Object Shape, Reflectance and Light Sources from Calibrated Images.
    B. Mercier, D. Meneveaux and A. Fournier.
    IJCV'07
    International Journal of Computer Vision IJCV, june 2007, Vol. 73, num. 1, pp.77-93.
  •        
  • Shape from Silhouette: Image Pixels for Marching Cubes.
    B. Mercier and D. Meneveaux.
    Journal of WSCG'05,
    The 13th International Conference in Central Europe on Computer Graphics, Visualization and ComputerVision, Vaclav Skala UNION Agency Ed., Vol.13, pp.112-118.
    Plzen, Czech Republic, Feb 2005.
  •        
  • Joint Estimation of Multiple Light Sources and Reflectance from Images.
    B. Mercier and D. Meneveaux.
    Proceedings of ICCVG'04,
    International Conference on Computer Vision and Graphics, Springer Verlag Ed., pp.66-71.
    Warsaw, Poland, Sep 2004.
  •        
  • Estimation conjointe de sources lumineuses et de propriétés photométriques à partir d'images.
    B. Mercier and D. Meneveaux.
    Proceedings of GTRV'04,
    Journées 2004 du Groupe de Travail Rendu et Visualisation (GDR ALP), pp.47-56.
    Poitiers, France, Apr 2004.
  •        
  • Lumigraphes et reconstruction géométrique.
    B. Mercier and D. Meneveaux.
    Proceedings of AFIG'03,
    16èmes Journées de l'Association Française d'Informatique Graphique, pp.241-252.
    Saint Denis, France, Dec 2003.





  • original image (photograph) voxel-based reconstruction (depth 7) triangular mesh (depth 7) surface reshaded with new light sources (using a lambertian reflectance model) chrominance-based classification (hue, saturation) surface reshaded with the detected light sources (using only estimated brdfs) lighting-based classification (number of light sources) surface reshaded with the detected light sources (using estimated brdfs and original images)