Ricardo Nakamura,  Romero Tori. 

In this paper we present a novel technique for the detection of collisions between a video avatar and entities of a virtual environment. This technique is suitable for real-time navigation and interaction of the video avatar within the virtual 3D space, and it is optimized to run on low-cost hardware platforms. It is based on the construction of an octree representation from a dense depth map and thus, takes into account the frame-to-frame posture changes from the video avatar. The results of a multithreaded implementation are presented and discussed.

http://www.lbd.dcc.ufmg.br/colecoes/svr/2008/0014.pdf

Processando consulta.