Video-Based Microfacet-Billboard Avatar for Educational Immersive Teleconference Systems

Daniel Makoto TokunagaSilvio R. R. SanchesLucas Padovani TriasJoão Luiz Bernardes JuniorRicardo NakamuraRomero Tori

Immersive teleconference systems have several applications,such as collaborative work, educationand entertainment, with different requirementsin terms of hardware and software. This paperpresents a video avatar system suitable to portable,commodity hardware-based educational systems.We extend the microfacet billboarding algorithmto provide good results without requiring a voxelor visual hull model, thus facilitating its use atinteractive frame rates. Our algorithm requires asingle image and a dense depth map for it, obtainedin this implementation with a zCam prototype.We also add normal mapping to the sys- tem,to allow the virtual lighting to affect the videoavatar. This rendering technique is combined withforeground segmentation, using a trimap imagemask, for instance, and head-tracking to provide anatural way for the user to control his viewpoint.The trade-o between image quality and reducedcosts enables such systems to be widely employedin distance learning where more sophisticatedimmersive systems would not be feasible. Thecharacteristics of our approach are detailed andcompared to related works in the field. Results of aprototype implementation used in preliminary usertests are also presented and discussed and show howthe system does indeed work at interactive framerates, requires low cost equipment and little setupand pro- vides a video avatar with a good sense ofimmersion in the virtual environment.

Caso o link acima esteja inválido, faça uma busca pelo texto completo na Web: Buscar na Web

Biblioteca Digital Brasileira de Computação - Contato:
     Mantida por: