Abstract: Possibilities for dynamic interactions of people with machines created by combination of virtual reality and communication networking provide new interesting problems at the intersection of two domains (among others): computer vision and computer graphics. In this paper, a technical solution to one of these problems is presented to automate the mixing of real and synthetic objects in a same animated video sequence. Current approaches usually involve mainly 2D-based effects and rely heavily on human expertise and interaction. We aim at achieving close interaction between 3D-based analysis and synthesis techniques to compute the interaction between a real scene captured in a sequence of calibrated images, and a computer-generated environment.